# MCP ToolChain — Full Directory (Markdown) > Every one of 155 verified MCP tools and vendors in the directory, in full, across 10 categories. Generated from the same data that backs https://mcptoolchain.ai/directory/. See https://mcptoolchain.ai/llms.txt for the index, and https://authplane.ai/ for AuthPlane, the MCP authorization product that maintains this directory. Last reviewed: 2026-08-21 ## Authorization & Identity Authorization servers and identity layers that issue and validate the tokens an MCP server accepts. The MCP authorization specification builds on OAuth 2.1, so this category covers both MCP-specific products and general-purpose authorization servers used for MCP. ### Auth0 for AI Agents - Vendor: Okta - Category: Authorization & Identity - Build stage: Secure access - Deployment: SaaS - Licence: Not applicable - Languages: JavaScript, TypeScript - Website: https://auth0.com/ai - Docs: https://auth0.com/ai/docs/intro/overview - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--auth0-for-genai/ Provides OAuth 2.1 and OpenID Connect authorization for MCP servers, including discovery, client registration, resource-scoped tokens, and token exchange. Why teams reach for it: Developers use it to add the MCP spec's OAuth 2.1 authorization to their server without building an authorization server themselves. Worth knowing: It implements OAuth 2.1, with PKCE required, and CIMD and DCR client registration. It federates with Google and holds an upstream token vault for third-party credentials. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Required - Token Exchange (RFC 8693): Yes - Idp Federation: Google - Upstream Token Vault: Yes --- ### AuthPlane - Vendor: AuthPlane - Category: Authorization & Identity - Build stage: Secure access - Deployment: Self-hosted - Licence: AGPL-3.0 - Languages: Go, TypeScript - Website: https://authplane.ai/ - Docs: https://docs.authplane.ai/ - Source: https://github.com/AuthPlane/authserver - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--authplane/ A self-hosted OAuth 2.1 authorization server for MCP servers. Why teams reach for it: Replaces per-MCP-server user databases, consent UX, and session models with a single authorization server for identity, authorization, audience binding, revocation, and audit. Worth knowing: Federates to Google Workspace, Okta, Entra ID, Auth0, or any OIDC provider rather than replacing them. SQLite for local use, PostgreSQL and Helm for production. A Token Vault holds per-user upstream credentials. Enterprise and cloud editions are stated as roadmap. This directory is built and maintained by AuthPlane; see the inclusion rule. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Required - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: OIDC - Upstream Token Vault: Yes - Self Hostable: Yes - Audit Log: Yes --- ### MCP - Vendor: Better Auth - Category: Authorization & Identity - Build stage: Secure access - Deployment: Self-hosted - Licence: MIT - Languages: TypeScript - Website: https://better-auth.com/docs/plugins/mcp - Source: https://github.com/better-auth/better-auth - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--better-auth-mcp-plugin/ Adds MCP provider authentication to an application and provides a framework-agnostic remote auth client for verifying tokens and protecting resources when the MCP server is separate from Better Auth. Why teams reach for it: Adds MCP provider authentication to your app. Worth knowing: MCP plugin will be deprecated in favor of the OAuth Provider Plugin Specifications: - Oauth Version: OAuth 2.1 - Client Registration: DCR only - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported - Upstream Token Vault: No - Self Hostable: Yes --- ### Clerk - Vendor: Clerk - Category: Authorization & Identity - Build stage: Secure access - Deployment: SaaS - Licence: Not applicable - Languages: JavaScript, C#, Go, Java, PHP, Python, Ruby - Website: https://clerk.com/ - Docs: https://clerk.com/docs - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--clerk/ Authentication platform with OAuth provider support used for MCP server authorization Why teams reach for it: It adds MCP capabilities to an existing application through one API endpoint without deploying and managing separate authenticated MCP services. Worth knowing: It with PKCE supported, and CIMD and DCR client registration. It federates with Google, Microsoft, Notion and holds an upstream token vault for third-party credentials. Specifications: - Client Registration: CIMD and DCR - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported - Idp Federation: Google, Microsoft, Notion - Upstream Token Vault: Yes - Audit Log: Yes --- ### Inbound Apps - Vendor: Descope - Category: Authorization & Identity - Build stage: Secure access - Deployment: SaaS - Licence: Not applicable - Website: https://www.descope.com/ - Docs: https://docs.descope.com/identity-federation/inbound-apps - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--descope/ Provides identity infrastructure for AI agents and MCP servers, including authorization, scoped access control, consent, token management, and DCR security. Why teams reach for it: Teams who want Descope to be the OAuth authorization server for their APIs and MCP servers, handling client registration, login and consent. Worth knowing: It implements OAuth 2.1, with PKCE supported, and CIMD and DCR client registration. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Pkce: Supported - Token Exchange (RFC 8693): Yes --- ### ext-auth - Vendor: Model Context Protocol - Category: Authorization & Identity - Build stage: Secure access - Deployment: Self-hosted - Licence: MIT - Languages: MDX - Website: https://github.com/modelcontextprotocol/ext-auth - Source: https://github.com/modelcontextprotocol/ext-auth - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--ext-auth/ Official repository of MCP authorization extensions Why teams reach for it: Adds modular MCP authorization mechanisms beyond the core protocol, including enterprise-managed SSO and machine-to-machine authentication without user interaction. Worth knowing: This repository follows the Model Context Protocol Governance process with dedicated maintainers for authorization extensions Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD only - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Required - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: Enterprise-Managed Authorization - Upstream Token Vault: No - Audit Log: No --- ### Keycloak - Vendor: Cloud Native Computing Foundation - Category: Authorization & Identity - Build stage: Secure access - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Java - Website: https://www.keycloak.org/ - Docs: https://www.keycloak.org/documentation - Source: https://github.com/keycloak/keycloak - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--keycloak/ Keycloak is open-source identity and access management for applications and services. Why teams reach for it: Adds authentication to applications and secures services with minimum effort, without requiring applications to store or authenticate users. Worth knowing: It implements OAuth 2.1, with PKCE supported, and CIMD and DCR client registration. It federates with OpenID Connect identity providers and holds an upstream token vault for third-party credentials and is self-hostable. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): No - Protected Resource Metadata (RFC 9728): No - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: OpenID Connect identity providers - Upstream Token Vault: Yes - Self Hostable: Yes - Audit Log: Yes --- ### Ory Hydra - Vendor: Ory - Category: Authorization & Identity - Build stage: Secure access - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Go - Website: https://www.ory.com/hydra - Docs: https://www.ory.com/docs/network/hydra - Source: https://github.com/ory/hydra - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--ory-hydra/ OpenID-certified OAuth 2.0 server that delegates user authentication to your application Why teams reach for it: Provides a standards-certified token issuer for teams that want to keep their own login experience and user store. Worth knowing: Not MCP-specific. Ory also offers a hosted network edition. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): No - Protected Resource Metadata (RFC 9728): No - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: any existing identity provider - Upstream Token Vault: No - Self Hostable: Yes - Audit Log: No --- ### Permit.io - Vendor: Permit.io - Category: Authorization & Identity - Build stage: Secure access - Deployment: Hybrid - Licence: MIT - Languages: Python - Website: https://www.permit.io/ - Docs: https://docs.permit.io/ - Source: https://github.com/permitio/permit-mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--permit-io/ Provides real-time, fine-grained authorization and policy-backed delegation, approvals, trust, and audit for humans, services, and AI agents. Why teams reach for it: Permit.io solves the problem of securing ephemeral AI agents by providing real-time, fine-grained authorization that replaces static permissions with policy-backed, just-in-time access. Worth knowing: The product requires admin grants for user access to MCP servers and has specific trust level configurations that affect user permissions. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD only - Authorization Server Metadata (RFC 8414): Yes - Pkce: Required - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: SAML 2.0 SSO, OIDC SSO - Upstream Token Vault: Yes - Self Hostable: Yes - Audit Log: Yes --- ### Prefactor - Vendor: Prefactor - Category: Authorization & Identity - Build stage: Secure access - Deployment: Hybrid - Licence: Not applicable - Languages: TypeScript, Python - Website: https://prefactor.tech/ - Docs: https://docs.prefactor.ai - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--prefactor/ Records agent activity, classifies data risk, and stores quality evidence for reviewing agents in production. Why teams reach for it: Makes the agent version running in each environment explicit and reviewable, including deliberate staging and production divergence. Specifications: - Self Hostable: Enterprise plan only --- ### Scalekit MCP Auth - Vendor: Scalekit - Category: Authorization & Identity - Build stage: Secure access - Deployment: SaaS - Licence: MIT - Languages: Node.js, Python, Go, Java - Website: https://www.scalekit.com/mcp-auth - Docs: https://docs.scalekit.com/authenticate/mcp/overview/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--scalekit/ Drop-in OAuth 2.1 for any MCP server — user login, SSO, agent tokens, scopes, and DCR included Why teams reach for it: Provides MCP-specific authorization endpoints without modifying an existing application's own auth stack. Worth knowing: It implements OAuth 2.1, with PKCE required, and CIMD and DCR client registration. It federates with Okta, Microsoft Entra ID, ADFS and holds an upstream token vault for third-party credentials and is self-hostable. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Required - Agent Delegation: Carried in token - Idp Federation: Okta, Microsoft Entra ID, ADFS - Upstream Token Vault: Yes - Self Hostable: Yes - Audit Log: Yes --- ### Stytch Connected Apps - Vendor: Stytch - Category: Authorization & Identity - Build stage: Secure access - Deployment: SaaS - Licence: Not applicable - Languages: Node, Go, Java, Kotlin, PHP, Python, Ruby, Rust - Website: https://stytch.com/ - Docs: https://stytch.com/docs/api-reference/b2b/api/oauth/token/google - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--stytch/ Provides AI-agent authentication, authorization, and consent management, including an MCP authentication and authorization toolkit. Why teams reach for it: Adds OAuth authorization for AI agents and MCP servers to an existing identity stack without replacing it. Worth knowing: It implements OAuth 2.0, with PKCE supported, and CIMD and DCR client registration. Specifications: - Oauth Version: OAuth 2.0 - Client Registration: CIMD and DCR - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported --- ### workers-oauth-provider - Vendor: Cloudflare - Category: Authorization & Identity - Build stage: Secure access - Deployment: Self-hosted - Licence: MIT - Languages: TypeScript - Website: https://github.com/cloudflare/workers-oauth-provider - Source: https://github.com/cloudflare/workers-oauth-provider - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--workers-oauth-provider/ Adds OAuth 2.1 to HTTP APIs and remote MCP servers on Workers; KV-backed, with Client ID Metadata Document support Why teams reach for it: Adds OAuth 2.1 authorization to HTTP APIs and remote MCP servers running on Cloudflare Workers. Worth knowing: External API key compatibility feature for MCP 2026-07-28 Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: https://idp.example.com - Upstream Token Vault: Yes - Self Hostable: Yes - Audit Log: No --- ### WorkOS - Vendor: WorkOS - Category: Authorization & Identity - Build stage: Secure access - Deployment: SaaS - Licence: MIT - Languages: TypeScript - Website: https://workos.com/mcp - Docs: https://workos.com/docs/authkit - Source: https://github.com/workos/mcp.shop - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--workos/ Provides OAuth authorization for MCP servers through AuthKit and WorkOS Connect. Why teams reach for it: Acts as the OAuth authorization server for MCP servers, whether AuthKit handles user login or the application already does. Worth knowing: It implements OAuth 2.1, with PKCE supported, and CIMD and DCR client registration. It federates with Okta, Entra ID, Auth0 SAML, Google OAuth, Microsoft OAuth, GitHub OAuth, Sign in with Apple, OpenID Connect, Custom SAML. Specifications: - Oauth Version: OAuth 2.1 - Client Registration: CIMD and DCR - Resource Indicators (RFC 8707): Yes - Protected Resource Metadata (RFC 9728): Yes - Authorization Server Metadata (RFC 8414): Yes - Pkce: Supported - Token Exchange (RFC 8693): Yes - Idp Federation: Okta, Entra ID, Auth0 SAML, Google OAuth, Microsoft OAuth, GitHub OAuth, Sign in with Apple, OpenID Connect, Custom SAML --- ### ZITADEL - Vendor: ZITADEL - Category: Authorization & Identity - Build stage: Secure access - Deployment: Hybrid - Licence: AGPL-3.0 - Languages: Go, TypeScript, MDX, JavaScript, PLpgSQL - Website: https://zitadel.com/ - Docs: https://zitadel.com/docs - Source: https://github.com/zitadel/zitadel - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/authorization-identity--zitadel/ Open-source identity and access management platform for SaaS and B2B applications, providing SSO, MFA, passkeys, OIDC, SAML, SCIM, and multi-tenancy with self-hosted and SaaS deployment options. Why teams reach for it: ZITADEL is an open-source identity and access management platform built for teams that need more than basic auth. Worth knowing: Postman Collection is in beta Specifications: - Oauth Version: OAuth 2.0 - Client Registration: DCR only - Resource Indicators (RFC 8707): No - Protected Resource Metadata (RFC 9728): No - Authorization Server Metadata (RFC 8414): No - Pkce: Required - Token Exchange (RFC 8693): Yes - Agent Delegation: Carried in token - Idp Federation: Entra ID, Google - Upstream Token Vault: No - Self Hostable: Yes - Audit Log: Yes --- ## Frameworks & SDKs Libraries you write code against to build an MCP server or client. Official protocol SDKs and the higher-level frameworks built on top of them. ### Apollo MCP Server - Vendor: Apollo GraphQL - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Self-hosted - Licence: MIT - Languages: Rust - Website: https://www.apollographql.com/apollo-mcp-server - Docs: https://www.apollographql.com/docs/apollo-mcp-server - Source: https://github.com/apollographql/apollo-mcp-server - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--apollo-mcp-server/ Translates GraphQL operations into MCP tools for AI applications and agents. Why teams reach for it: Exposes GraphQL APIs to AI agents without requiring code changes or custom integrations. Worth knowing: It implements MCP specification revision 2025-11-25, over stdio, Streamable HTTP. Specifications: - Language: Rust - Official Sdk: No - Spec Revision Implemented: 2025-11-25 - Transports: stdio, Streamable HTTP - Auth Helpers: Built-in - Mcp Apps Support: Yes - Package Registry: Not applicable - Package Name: apollo-mcp-server --- ### FastAPI-MCP - Vendor: Tadata - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Python - Website: https://fastapi-mcp.tadata.com/ - Source: https://github.com/tadata-org/fastapi_mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--fastapi-mcp/ Converts FastAPI applications into Model Context Protocol servers with built-in authentication support. Exposes API endpoints as MCP tools without external HTTP overhead. Why teams reach for it: Eliminates HTTP overhead and external communication when exposing FastAPI services to AI agents through MCP. Worth knowing: It implements MCP specification revision 2025-03-26, builds servers, over SSE, Streamable HTTP, and can generate tools directly from an OpenAPI spec. Specifications: - Language: Python - Official Sdk: No - Spec Revision Implemented: 2025-03-26 - Transports: SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: No - Auth Helpers: Built-in - Generates From Openapi: Yes - Mcp Apps Support: No - Server Composition: No - Package Registry: PyPI - Package Name: fastapi-mcp --- ### FastMCP - Vendor: Prefect - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: Apache-2.0 - Languages: Python - Website: https://gofastmcp.com/ - Source: https://github.com/jlowin/fastmcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--fastmcp/ FastMCP is a full framework for building MCP applications, providing unified APIs for servers, clients, and interactive apps. Why teams reach for it: Simplifies MCP development with one coherent API for servers, clients, and apps. Worth knowing: It implements MCP specification revision 2026-07-28, builds servers and clients, over stdio, SSE, Streamable HTTP, and can generate tools directly from an OpenAPI spec. Specifications: - Language: Python - Official Sdk: No - Spec Revision Implemented: 2026-07-28 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: Yes - Mcp Apps Support: Yes - Server Composition: Yes - Package Registry: PyPI - Package Name: fastmcp --- ### foxy-contexts - Vendor: strowk - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Go - Website: https://github.com/strowk/foxy-contexts - Source: https://github.com/strowk/foxy-contexts - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--foxy-contexts/ Foxy Contexts is a Go library for building MCP context servers using a declarative approach with dependency injection. Why teams reach for it: foxy-contexts builds Go MCP servers on top of uber's fx dependency-injection container, so each tool, resource or prompt is declared next to the code that serves it and registered with an app.Builder while shared clients and database connections are injected rather than threaded through by hand, and it ships a foxytest package for functional tests of the finished server. Specifications: - Language: Go - Official Sdk: No - Spec Revision Implemented: 2025-03-26 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Package Registry: Go modules - Package Name: github.com/strowk/foxy-contexts --- ### hyper-mcp - Vendor: hyper-mcp-rs - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Rust - Website: https://github.com/hyper-mcp-rs/hyper-mcp - Source: https://github.com/hyper-mcp-rs/hyper-mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--hyper-mcp/ hyper-mcp is an MCP server whose tools come from WebAssembly plugins, which can be written in any language that compiles to WebAssembly and distributed through OCI container registries. Plugins run sandboxed with configurable limits on network, filesystem and memory access, and plugin images pulled from OCI registries have their signatures verified at load time. Why teams reach for it: Its tools are WebAssembly plugins rather than trusted local processes, so each one can be written in any language that compiles to Wasm, shipped through an ordinary OCI registry, and sandboxed with limits on its network, filesystem and memory access. Specifications: - Language: Rust - Official Sdk: No - Transports: stdio - Builds Servers: Yes - Builds Clients: No - Auth Helpers: None - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: crates.io - Package Name: hyper-mcp --- ### langchain-mcp-adapters - Vendor: LangChain - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Python - Website: https://github.com/langchain-ai/langchain-mcp-adapters - Source: https://github.com/langchain-ai/langchain-mcp-adapters - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--langchain-mcp-adapters/ This library provides a lightweight wrapper that makes Model Context Protocol (MCP) tools compatible with LangChain and LangGraph. Why teams reach for it: It is the first-party bridge between MCP and LangChain/LangGraph: it converts the tools an MCP server exposes into LangChain tools an agent can call directly, and its client loads them from several MCP servers at once, so no per-server wrapper has to be written. Specifications: - Language: Python - Official Sdk: No - Spec Revision Implemented: 2025-06-18 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: No - Builds Clients: Yes - Auth Helpers: None - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: PyPI - Package Name: langchain-mcp-adapters --- ### mcp-agent - Vendor: LastMile AI - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Python - Website: https://github.com/lastmile-ai/mcp-agent - Source: https://github.com/lastmile-ai/mcp-agent - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-agent/ mcp-agent is a framework for building AI agents from capabilities exposed by Model Context Protocol servers, handling the mechanics of connecting to servers, working with LLMs, handling external signals such as human input, and persisting state via durable execution. Applications built with it can themselves be exposed as MCP servers. Why teams reach for it: mcp-agent manages the lifecycle of MCP server connections for you and implements the patterns from Anthropic's Building Effective Agents - router, orchestrator, evaluator-optimizer, map-reduce - as composable pieces, and the same agent code scales onto Temporal as a durable runtime, gaining pause, resume and recovery without API changes. Specifications: - Language: Python - Official Sdk: No - Spec Revision Implemented: 2025-06-18 - Transports: stdio, SSE, Streamable HTTP - Auth Helpers: Built-in - Package Registry: PyPI - Package Name: mcp-agent --- ### MCP C# SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: C# - Website: https://github.com/modelcontextprotocol/csharp-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/csharp-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-csharp-sdk/ The official C# SDK for the Model Context Protocol, enabling .NET applications to implement MCP clients and servers. Why teams reach for it: A .NET team gets the official MCP implementation, maintained with Microsoft, packaged to match how the application is hosted: a dependency-light Core package for client or low-level server use, a main package carrying the hosting and dependency-injection extensions, and a separate ASP.NET Core package for HTTP-based servers. Specifications: - Language: C# - Official Sdk: Yes - Spec Revision Implemented: 2026-07-28 - Transports: SSE, Streamable HTTP, stdio - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: Yes - Server Composition: No - Package Registry: NuGet - Package Name: ModelContextProtocol --- ### mcp-framework - Vendor: QuantGeekDev - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: TypeScript - Website: https://github.com/QuantGeekDev/mcp-framework - Source: https://github.com/QuantGeekDev/mcp-framework - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-framework/ MCP-Framework is a TypeScript framework for building Model Context Protocol servers. It discovers tools, resources and prompts from the project directory automatically, provides base classes for defining them, and serves them over stdio, SSE or HTTP Stream. Why teams reach for it: It removes the per-server boilerplate of the official SDK: tools, resources and prompts are discovered from the project directory and declared as typed base classes, and the transports and SSE authentication are already wired. Specifications: - Language: TypeScript - Official Sdk: No - Spec Revision Implemented: 2025-11-25 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: No - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: Yes - Server Composition: No - Package Registry: npm - Package Name: mcp-framework --- ### mcp-go - Vendor: mark3labs - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Go, MDX - Website: https://github.com/mark3labs/mcp-go - Docs: https://mcp-go.dev/ - Source: https://github.com/mark3labs/mcp-go - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-go/ mcp-go is a Go library for building MCP servers and clients: a server is created with `server.NewMCPServer`, tools are declared with typed parameter helpers and registered with `AddTool`, and the library handles connection management, protocol compliance and message routing. It implements the MCP specification version 2025-11-25 with backward compatibility to 2024-11-05, and runs over stdio, SSE, StreamableHTTP or an in-process transport. Why teams reach for it: Go developers need a library for building Model Context Protocol servers with minimal boilerplate. Specifications: - Language: Go - Official Sdk: No - Spec Revision Implemented: 2026-07-28 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Package Registry: Go modules - Package Name: github.com/mark3labs/mcp-go --- ### MCP Go SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Go - Website: https://github.com/modelcontextprotocol/go-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/go-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-go-sdk/ A library that provides Go APIs for constructing and using MCP clients and servers. Why teams reach for it: It is the official Model Context Protocol SDK for Go, published in the modelcontextprotocol organisation and aiming to implement the full MCP spec, where the established Go alternatives (mcp-go, mcp-golang, go-mcp) are third-party projects. Specifications: - Language: Go - Official Sdk: Yes - Spec Revision Implemented: 2026-07-28 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: Go modules - Package Name: github.com/modelcontextprotocol/go-sdk --- ### mcp-handler - Vendor: Vercel - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: Apache-2.0 - Languages: TypeScript, JavaScript - Website: https://github.com/vercel/mcp-handler - Source: https://github.com/vercel/mcp-handler - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-handler/ mcp-handler is a framework-agnostic HTTP adapter for hosting Model Context Protocol (MCP) servers in JavaScript and TypeScript applications. Why teams reach for it: It is a framework-agnostic HTTP adapter that turns an MCP server definition into a Web-standard (Request) => Promise handler, so the same MCP server can be mounted in Next.js, Nuxt/Nitro, SvelteKit, Hono or any other Fetch-compatible framework. Specifications: - Language: TypeScript - Official Sdk: No - Spec Revision Implemented: 2026-07-28 - Transports: Streamable HTTP - Builds Servers: Yes - Builds Clients: No - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: npm - Package Name: mcp-handler --- ### MCP Java SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Java - Website: https://github.com/modelcontextprotocol/java-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/java-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-java-sdk/ A Java SDK for building MCP clients and servers with support for tools, resources, prompts, sampling, and multiple protocol versions using reactive and blocking APIs. Why teams reach for it: Enables Java developers to quickly build MCP clients and servers using familiar, widely-adopted libraries with flexible, pluggable components. Worth knowing: It implements MCP specification revision 2025-11-25, builds servers and clients, over stdio, SSE, Streamable HTTP. Specifications: - Language: Java - Official Sdk: Yes - Spec Revision Implemented: 2025-11-25 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: None - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: Maven Central - Package Name: mcp-parent --- ### MCP Kotlin SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Kotlin - Website: https://github.com/modelcontextprotocol/kotlin-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/kotlin-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-kotlin-sdk/ Kotlin SDK implementing the Model Context Protocol specification. Enables building MCP clients and servers targeting JVM, Native, JS, and Wasm platforms using a single codebase. Why teams reach for it: A Kotlin team gets the official SDK, maintained with JetBrains, as a Kotlin Multiplatform library: one codebase of client and server code targets JVM, Native, JS and Wasm over stdio, SSE, Streamable HTTP and WebSocket, the APIs are coroutine-based, and client-only or server-only artifacts can be taken instead of the umbrella one. Specifications: - Language: Kotlin - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Package Registry: Maven Central - Package Name: io.modelcontextprotocol:kotlin-sdk --- ### MCP PHP SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: PHP - Website: https://github.com/modelcontextprotocol/php-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/php-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-php-sdk/ The official PHP SDK for Model Context Protocol (MCP). It provides a framework-agnostic API for implementing MCP servers and clients in PHP. Why teams reach for it: It is the official PHP SDK for MCP - a PHP Foundation and Symfony collaboration - giving PHP applications one framework-agnostic API for both MCP servers and MCP clients. Specifications: - Language: PHP - Official Sdk: Yes - Spec Revision Implemented: 2026-07-28 - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: Yes - Server Composition: No - Package Registry: Packagist - Package Name: mcp/sdk --- ### MCP Python SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Python - Website: https://github.com/modelcontextprotocol/python-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/python-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-python-sdk/ The official Python SDK for the Model Context Protocol, used to build MCP servers that expose tools, resources and prompts and MCP clients that connect to them. Its v2 release line implements the 2026-07-28 protocol revision and answers both 2025-era and 2026-era clients from the same stdio or Streamable HTTP server. Why teams reach for it: It is the official Python SDK for MCP and covers both sides of the protocol in one package: a server is two type-hinted Python functions with no JSON Schema, parsing, validation or protocol code to write, and the same package is a full client over stdio, Streamable HTTP and SSE. Specifications: - Language: Python - Official Sdk: Yes - Spec Revision Implemented: 2026-07-28 - Transports: SSE, Streamable HTTP, stdio - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: Yes - Server Composition: No - Package Registry: PyPI - Package Name: mcp --- ### MCP Ruby SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Ruby - Website: https://github.com/modelcontextprotocol/ruby-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/ruby-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-ruby-sdk/ The official Ruby SDK for Model Context Protocol servers and clients. Why teams reach for it: The official Ruby SDK for the Model Context Protocol, implementing both server and client functionality for JSON-RPC 2.0 communication between LLM applications and context providers - the first-party option for putting MCP into a Ruby or Rails application. Specifications: - Language: Ruby - Official Sdk: Yes - Spec Revision Implemented: 2026-07-28 - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: None - Generates From Openapi: No - Mcp Apps Support: Yes - Server Composition: No - Package Registry: RubyGems - Package Name: mcp --- ### RMCP - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Rust - Website: https://github.com/modelcontextprotocol/rust-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/rust-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-rust-sdk/ An official Rust implementation of the Model Context Protocol SDK that supports building both servers and clients using tokio async runtime. Why teams reach for it: For a Rust service this is the official SDK and it tracks the specification closely: it implements the stable 2026-07-28 revision, including server discovery and negotiation, transport-neutral subscriptions, long-running tasks, response caching, multi-round-trip requests and standard HTTP routing headers, while remaining compatible with 2025-11-25 and earlier peers, and the same rmcp crate builds either end of the connection on the tokio runtime. Specifications: - Language: Rust - Spec Revision Implemented: 2026-07-28 - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Package Registry: crates.io - Package Name: rmcp --- ### MCP Swift SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Swift - Website: https://github.com/modelcontextprotocol/swift-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/swift-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-swift-sdk/ Official Swift SDK implementing both MCP client and server components for standardized communication with AI and ML models. Why teams reach for it: The official Swift SDK for the Model Context Protocol, implementing both client and server components against the current specification revision, so Swift and Apple-platform applications get first-party MCP support. Specifications: - Language: Swift - Official Sdk: Yes - Spec Revision Implemented: 2025-11-25 - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: Not applicable - Package Name: MCP - the Swift Package Manager library product of https://github.com/modelcontextprotocol/swift-sdk.git --- ### MCP Toolbox for Databases - Vendor: Google - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Go - Website: https://github.com/googleapis/mcp-toolbox - Docs: https://googleapis.github.io/mcp-toolbox/ - Source: https://github.com/googleapis/mcp-toolbox - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-toolbox-for-databases/ An open source MCP server connecting AI agents and applications directly to enterprise databases with prebuilt tools for data exploration and a framework for building secure custom tools. Why teams reach for it: One governed MCP server between your agents, IDEs and enterprise databases: prebuilt generic database tools out of the box plus a framework for custom tools, with connection pooling, IAM auth and OpenTelemetry handled centrally so tools can be updated and shared without redeploying every application. Specifications: - Language: Go - Official Sdk: No - Spec Revision Implemented: 2026-07-28 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: Go modules - Package Name: github.com/googleapis/mcp-toolbox --- ### MCP TypeScript SDK - Vendor: Model Context Protocol - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: TypeScript, JavaScript - Website: https://github.com/modelcontextprotocol/typescript-sdk - Docs: https://modelcontextprotocol.io/docs/sdk - Source: https://github.com/modelcontextprotocol/typescript-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-typescript-sdk/ The TypeScript SDK implementation of the Model Context Protocol. Provides libraries for building MCP servers and clients. Why teams reach for it: The official TypeScript implementation of the MCP specification, shipping first-party server and client libraries plus runtime and framework middleware, so a Node.js, Bun or Deno application does not hand-roll the protocol. Specifications: - Language: TypeScript - Official Sdk: Yes - Spec Revision Implemented: 2026-07-28 - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: No - Mcp Apps Support: No - Server Composition: No - Package Registry: npm - Package Name: @modelcontextprotocol/sdk --- ### MCP-UI - Vendor: Ido Salomon - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: Apache-2.0 - Languages: TypeScript, Python - Website: https://mcpui.dev/ - Docs: https://mcpui.dev/guide/introduction - Source: https://github.com/MCP-UI-Org/mcp-ui - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-ui/ MCP-UI is a set of SDK packages that let an MCP tool return an interactive HTML interface instead of plain text, and let the host render it. It implements the MCP Apps standard: on the server side a tool is linked to a UI resource through _meta.ui.resourceUri, and on the client side AppRenderer fetches that resource and renders it in a sandboxed iframe. Why teams reach for it: You reach for MCP-UI when you want both ends of an MCP Apps integration from one project: server SDKs in TypeScript, Python and Ruby that create the UI resource, React client components that render it, and an adapter that lets widgets written against the older MCP-UI postMessage protocol run inside MCP Apps hosts without code changes. Specifications: - Language: TypeScript - Official Sdk: No - Transports: Not applicable - Mcp Apps Support: Yes - Package Registry: npm - Package Name: mcp-ui --- ### mcp-use - Vendor: Manufact - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: Python, TypeScript - Website: https://mcp-use.com/ - Docs: https://docs.mcp-use.com/ - Source: https://github.com/mcp-use/mcp-use - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--mcp-use/ Full-stack MCP framework for TypeScript and Python. Build MCP servers, clients, and agents with support for interactive Views in ChatGPT and Claude. Why teams reach for it: Bundles server framework, development server, Inspector, and cloud deployment to reduce setup complexity. Worth knowing: It implements MCP specification revision 2026-07-28, builds servers and clients, over stdio, SSE, Streamable HTTP, and can generate tools directly from an OpenAPI spec. Specifications: - Language: TypeScript - Official Sdk: No - Spec Revision Implemented: 2026-07-28 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Built-in - Generates From Openapi: Yes - Mcp Apps Support: Yes - Server Composition: Yes - Package Registry: npm - Package Name: mcp-use --- ### openapi-to-mcpserver - Vendor: Higress - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: CLI tool - Licence: Apache-2.0 - Languages: Go - Website: https://github.com/higress-group/openapi-to-mcpserver - Source: https://github.com/higress-group/openapi-to-mcpserver - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--openapi-to-mcpserver/ openapi-to-mcpserver is a Go command-line tool that reads an OpenAPI specification in JSON or YAML and writes an MCP server configuration file, converting each OpenAPI path into a tool with its arguments, request template and response template. The output is a configuration rather than server code: it is added to a Higress gateway configuration, and a template file can patch common headers, authentication or output schemas into every tool. Why teams reach for it: It removes the hand-translation between an existing REST API and an MCP tool list: one command turns every path into a tool, sets each path, query, header, cookie and body parameter to its right position, carries the OpenAPI descriptions and security schemes across, and emits output schemas for MCP 2025-06-18 structured content, so the API reaches agents through Higress from its specification alone. Specifications: - Language: Go - Official Sdk: No - Spec Revision Implemented: 2025-06-18 - Transports: Not applicable - Generates From Openapi: Yes - Package Registry: Go modules - Package Name: github.com/higress-group/openapi-to-mcpserver --- ### Quarkus MCP Server - Vendor: Quarkiverse - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: Apache-2.0 - Languages: Java - Website: https://quarkus.io/extensions/io.quarkiverse.mcp/quarkus-mcp-server-sse/ - Source: https://github.com/quarkiverse/quarkus-mcp-server - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--quarkus-mcp-server/ Quarkus MCP Server is a Quarkus extension that provides declarative and programmatic APIs for implementing Model Context Protocol server features in Java. Server features - prompts, resources and tools - are written as annotated methods of CDI beans. Why teams reach for it: It lets a Quarkus application expose MCP tools, prompts and resources as annotated methods on its own CDI beans, so an existing Java service becomes an MCP server without a separate codebase. Specifications: - Language: Java - Official Sdk: No - Spec Revision Implemented: 2026-07-28 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: No - Auth Helpers: Separate package - Package Registry: Maven Central - Package Name: quarkiverse-parent --- ### Skybridge - Vendor: Alpic - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: TypeScript - Website: https://github.com/alpic-ai/skybridge - Source: https://github.com/alpic-ai/skybridge - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--skybridge/ Skybridge is a React and TypeScript framework for building MCP servers whose tools return interactive UI views that render inside MCP clients such as Claude and ChatGPT. It ships a dev server with a local emulator, hot module reload and a tunnel that connects a local app to those hosts. Why teams reach for it: Skybridge is for MCP servers that also have to render a UI: it supplies React hooks and tRPC-style type inference running from the MCP server's tool definition through to the React view, a dev server with a local emulator, hot module reload and a permanent tunnel into Claude and ChatGPT, and it abstracts the implementation differences between MCP clients so one app runs in Claude, ChatGPT and VSCode. Specifications: - Language: TypeScript - Official Sdk: No - Transports: Streamable HTTP - Builds Servers: Yes - Auth Helpers: Built-in - Mcp Apps Support: Yes - Package Registry: npm - Package Name: @skybridge/monorepo --- ### Speakeasy - Vendor: Speakeasy - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: SaaS - Licence: Elastic-2.0 - Languages: JavaScript, Go - Website: https://www.speakeasy.com/ - Docs: https://www.speakeasy.com/docs - Source: https://github.com/speakeasy-api/speakeasy - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--speakeasy/ Speakeasy generates MCP servers from an OpenAPI document, producing a TypeScript server whose tools map to the API's operations. It also runs a hosted control plane that builds, proxies and governs MCP servers for an organisation. Why teams reach for it: It turns an API's existing OpenAPI document into a deployable MCP server, so an API team gets one without hand-writing tools. Specifications: - Language: TypeScript - Official Sdk: No - Transports: Streamable HTTP - Builds Servers: Yes - Auth Helpers: Built-in - Generates From Openapi: Yes - Server Composition: Yes - Package Registry: Go modules - Package Name: github.com/speakeasy-api/speakeasy --- ### Spring AI MCP - Vendor: Spring - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: Apache-2.0 - Languages: Java - Website: https://docs.spring.io/spring-ai/reference/api/mcp/mcp-overview.html - Source: https://github.com/spring-projects/spring-ai - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--spring-ai-mcp/ Spring AI MCP provides Spring Boot starters and Java annotations for building Model Context Protocol servers and clients on top of the MCP Java SDK. A server application selects the STDIO, SSE, or Streamable HTTP protocol through Spring Boot configuration properties. Why teams reach for it: A Spring developer can add MCP to an existing Spring Boot application through Boot starters and Java annotations, on both the server and the client side, without leaving the Spring programming model. Specifications: - Language: Java - Official Sdk: No - Spec Revision Implemented: 2025-06-18 - Transports: stdio, SSE, Streamable HTTP - Builds Servers: Yes - Builds Clients: Yes - Auth Helpers: Separate package - Package Registry: Maven Central - Package Name: spring-ai-parent --- ### Stainless - Vendor: Stainless - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://www.stainless.com/ - Docs: https://www.stainless.com/docs - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--stainless/ Stainless automatically generates SDKs and MCP servers from an OpenAPI specification. Its generated MCP servers use a two-tool, code-execution architecture instead of exposing one tool per API method. Why teams reach for it: Teams that already maintain a traditional SDK use Stainless to add an MCP server without keeping a second codebase in sync by hand. Specifications: - Language: TypeScript - Official Sdk: No - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Auth Helpers: Built-in - Generates From Openapi: Yes - Package Registry: npm --- ### xmcp - Vendor: Basement Studio - Category: Frameworks & SDKs - Build stage: Choose a framework - Deployment: Library - Licence: MIT - Languages: TypeScript - Website: https://xmcp.dev/ - Docs: https://xmcp.dev/docs - Source: https://github.com/basementstudio/xmcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/frameworks-sdks--xmcp/ xmcp is a TypeScript framework for building and shipping MCP servers. Tools, prompts and resources are registered automatically from the file system, and a server can be built with either the HTTP or the STDIO transport. Why teams reach for it: xmcp turns an MCP server into a directory of TypeScript files: a file placed in the tools, prompts or resources directory is discovered and registered automatically, HTTP and STDIO transports are built in, and one command plugs the server into an existing Next.js or Express app. Specifications: - Language: TypeScript - Official Sdk: No - Spec Revision Implemented: 2025-11-25 - Transports: stdio, Streamable HTTP - Builds Servers: Yes - Package Registry: npm - Package Name: xmcp-monorepo --- ## Security & Governance Products that discover, assess, govern, or protect MCP servers, tools, agents, and their traffic. This category includes static scanners, runtime guardrails, posture management, threat detection, data-loss prevention, and access-governance products when security is their primary function. ### Proofpoint AI MCP Security - Vendor: Proofpoint - Category: Security & Governance - Build stage: Govern & protect - Deployment: Not disclosed - Licence: Apache-2.0 - Languages: Go - Website: https://www.proofpoint.com/us/products/ai-mcp-security - Source: https://github.com/acuvity/minibridge - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--acuvity/ Proofpoint AI MCP Security gives you the visibility, policy enforcement, and protection to govern every MCP server in your organization. Why teams reach for it: It solves the problem of unsecured and ungoverned MCP servers by automating discovery, classifying risk, enforcing policies, and routing traffic through secure gateways. Worth knowing: scanning network traffic. It supports shadow MCP discovery. Specifications: - Deployment Point: In request path - Scan Target: network traffic - Shadow Mcp Discovery: Yes - Clients Covered: Not client-specific - Policy Enforcement: Both - Siem Integration: Yes --- ### Akto - Vendor: Akto - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: MIT - Languages: Java, JavaScript, Python, Vue, Go, TypeScript, Lua - Website: https://www.akto.io/ - Docs: https://docs.akto.io/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--akto/ Akto maps every AI agent and MCP tool, gives visibility, runs continuous red teaming, and enforces guardrails at scale. Why teams reach for it: It combines discovery, continuous red teaming, and guardrail enforcement for AI agents and MCP tools. Worth knowing: Its primary function is static scanning, scanning MCP server config, tool descriptions, agent skills, among other targets. It supports shadow MCP discovery, prompt-injection detection, tool-poisoning detection. Specifications: - Primary Function: Static scanning - Secondary Functions: Runtime guardrails, Posture management, Threat detection - Deployment Point: In request path - Scan Target: MCP server config, tool descriptions, agent skills, network traffic, source code, running processes - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Amp, Claude Code CLI, Claude Agent SDK, Kiro CLI - Policy Enforcement: Both - Siem Integration: Yes - Compliance Frameworks: OWASP --- ### Barndoor - Vendor: Barndoor AI - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: Not applicable - Languages: Python, TypeScript - Website: https://barndoor.ai/ - Docs: https://docs.barndoor.ai - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--barndoor-ai/ Barndoor provides an LLM gateway that governs AI requests. It applies vendor-specific cache instructions, data protection, and request measurement. Why teams reach for it: It reduces the engineering work and cost uncertainty involved in maintaining vendor-specific prompt caching, security, routing, and monitoring. Worth knowing: Its primary function is access governance, scanning tool descriptions. It supports prompt-injection detection. Specifications: - Primary Function: Access governance - Secondary Functions: Runtime guardrails, Data loss prevention - Deployment Point: In request path - Scan Target: tool descriptions - Prompt Injection Detection: Yes - Clients Covered: Not client-specific - Policy Enforcement: Blocking - Siem Integration: Yes --- ### BlueRock - Vendor: BlueRock - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: Not applicable - Languages: Python - Website: https://www.bluerock.io/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--bluerock/ BlueRock traces agent actions across tools and MCP servers and applies runtime controls at execution. Why teams reach for it: Applies runtime guardrails to agent and MCP tool calls without requiring developers to change their frameworks, IDEs, or tooling. Worth knowing: BlueRock's MCP Trust Registry scans servers against 22 rules tied to OWASP MCP Top 10, MAESTRO, and MITRE CWE. It classifies tools as read, write, destructive, admin, iam, or cost-impact. For internal builds, private repo scanning covers servers that will never be public. BlueRock Guardrails enforce at the app runtime, blocking privilege escalation, SSRF, credential harvesting, and CLI-wrapper exploits before execution, at under 5ms overhead. Specifications: - Scan Target: source code - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Claude Code, Cursor, GitHub Copilot, Codex, VS Code - Policy Enforcement: Blocking - Siem Integration: Yes - Compliance Frameworks: OWASP --- ### Enkrypt AI - Vendor: Enkrypt AI - Category: Security & Governance - Build stage: Govern & protect - Deployment: SaaS - Licence: Apache-2.0 - Languages: Python, Jupyter Notebook - Website: https://www.enkryptai.com/ - Docs: https://docs.enkryptai.com/home/introduction - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--enkrypt-ai/ Runs a secure MCP gateway with authentication, automatic tool discovery, caching, and guardrail enforcement. Why teams reach for it: It puts a gateway between your MCP client and your MCP servers that enforces per-server guardrails on both the input and the output side of every tool call, backed by the same platform's MCP vulnerability scanning of repos, packages and live endpoints. Specifications: - Primary Function: Static scanning - Secondary Functions: Runtime guardrails - Deployment Point: In request path - Scan Target: MCP server config, tool descriptions, agent skills, network traffic, source code, running processes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Not client-specific - Policy Enforcement: Blocking - Siem Integration: Yes - Compliance Frameworks: NIST AI RMF, MITRE ATLAS, OWASP Top 10 for LLMs, EU AI Act, ISO/IEC 42001, ISO/IEC 27001 --- ### Lasso Security - Vendor: Lasso Security - Category: Security & Governance - Build stage: Govern & protect - Deployment: Self-hosted - Licence: MIT - Languages: Not applicable - Website: https://www.lasso.security/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--lasso-security/ Runs an intermediary gateway that proxies requests to MCP servers and manages their lifecycle. Why teams reach for it: Governs what an agent is trying to do across its whole execution chain, rather than filtering prompt and response content. Worth knowing: scanning tool descriptions. It supports shadow MCP discovery, prompt-injection detection, tool-poisoning detection. Specifications: - Secondary Functions: Runtime guardrails, Posture management, Threat detection, Data loss prevention, Access governance - Deployment Point: In request path - Scan Target: tool descriptions - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Claude Desktop, Claude Code, Cursor, Windsurf - Policy Enforcement: Both - Siem Integration: Yes - Compliance Frameworks: NIST AI Risk Management Framework, SOC 2 Type 2, ISO 27001, ISO 42001, EU AI Act, HIPAA, PCI-DSS, FedRAMP --- ### MCP Manager - Vendor: Usercentrics - Category: Security & Governance - Build stage: Govern & protect - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://mcpmanager.ai/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--mcp-manager/ Provides MCP gateways with observability, security rules, OAuth and RBAC enforcement, and server and tool governance. Why teams reach for it: Pins each server's approved tool schema and blocks changed tools until reviewed and re-approved, guarding against rug pull attacks. Worth knowing: scanning tool descriptions, network traffic. It supports prompt-injection detection, tool-poisoning detection. Specifications: - Secondary Functions: Data loss prevention, Runtime guardrails, Threat detection, Access governance - Deployment Point: In request path - Scan Target: tool descriptions, network traffic - Shadow Mcp Discovery: No - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Not client-specific - Policy Enforcement: Both - Siem Integration: Yes --- ### Snyk Agent Scan - Vendor: Snyk - Category: Security & Governance - Build stage: Govern & protect - Deployment: CLI tool - Licence: Apache-2.0 - Languages: Python - Website: https://github.com/snyk/agent-scan - Docs: https://github.com/snyk/agent-scan/blob/main/docs/README.md - Source: https://github.com/snyk/agent-scan - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--mcp-scan/ Discovers and scans MCP servers, tools, prompts, resources, skills, and supported agent configurations for security risks. Why teams reach for it: Discovers and scans the MCP servers and agent skills already configured on a machine, across many agent clients and config scopes. Worth knowing: Its primary function is static scanning, scanning MCP server config, tool descriptions, agent skills. It supports shadow MCP discovery, prompt-injection detection, tool-poisoning detection. Specifications: - Primary Function: Static scanning - Secondary Functions: Runtime guardrails, Posture management - Deployment Point: CLI or CI - Scan Target: MCP server config, tool descriptions, agent skills - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Windsurf, Cursor, VS Code, Claude Desktop, Claude Code, Gemini CLI, OpenClaw, Amp, Kiro, OpenCode, Antigravity, Codex, Amazon Q - Policy Enforcement: Alerting only - Siem Integration: Yes --- ### Noma Security - Vendor: Noma Security - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: Not applicable - Languages: Python, JavaScript - Website: https://noma.security/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--noma-security/ Noma Security provides continuous governance of agent Capabilities, Autonomy, and Permissions across AI environments, enabling visibility and protection of AI assets from development to production. Why teams reach for it: Covers discovery, pre-deployment policy and runtime enforcement for models, agents and MCP servers in one platform, without architecture changes. Worth knowing: Noma Security provides visibility and protection for MCP activity, reducing risk and strengthening governance. The platform includes AI Security Posture Management (AI-SPM) for continuous monitoring, compliance, and threat detection. Specifications: - Secondary Functions: Posture management, Runtime guardrails, Access governance, Threat detection, Data loss prevention - Deployment Point: In request path - Scan Target: network traffic - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Clients Covered: Cursor, Windsurf - Policy Enforcement: Blocking - Compliance Frameworks: OWASP, MITRE ATLAS, NIST --- ### Operant AI - Vendor: Operant AI - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: Not applicable - Website: https://www.operant.ai/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--operant-ai/ Runtime protection and visibility across an MCP estate Why teams reach for it: Blocks and redacts sensitive AI and agent data flows in-line at runtime, rather than reporting them for later review. Worth knowing: Its primary function is runtime guardrails, scanning network traffic, agent skills, running processes. It supports shadow MCP discovery, prompt-injection detection, tool-poisoning detection. Specifications: - Primary Function: Runtime guardrails - Secondary Functions: Threat detection, Data loss prevention, Access governance, Posture management - Scan Target: network traffic, agent skills, running processes - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Claude Code, Claude CLI, Claude Cowork - Policy Enforcement: Both - Siem Integration: Yes - Compliance Frameworks: OWASP LLM Top 10, NIST AI RMF, SOC 2, HIPAA --- ### Pomerium - Vendor: Pomerium - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Go, TypeScript, JavaScript, Jsonnet, Lua - Website: https://www.pomerium.com/ - Docs: https://www.pomerium.com/docs - Source: https://github.com/pomerium/pomerium - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--pomerium/ Pomerium is an identity and context-aware reverse proxy. It builds clientless connections to internal web apps and other services without a corporate VPN. Why teams reach for it: Pomerium builds secure, clientless connections to internal web apps and other services without a corporate VPN. Worth knowing: Pomerium requires a persistent databroker backend for caching tokens and upstream OAuth state; it does not support in-memory backend for production use. Specifications: - Primary Function: Access governance - Secondary Functions: None - Deployment Point: In request path - Scan Target: Not applicable - Shadow Mcp Discovery: No - Prompt Injection Detection: No - Tool Poisoning Detection: No - Clients Covered: Not client-specific - Policy Enforcement: Blocking - Siem Integration: Yes - Compliance Frameworks: Not applicable --- ### Runlayer - Vendor: Runlayer - Category: Security & Governance - Build stage: Govern & protect - Deployment: Hybrid - Licence: Not applicable - Languages: TypeScript, Python - Website: https://www.runlayer.com/ - Docs: https://docs.runlayer.com - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--runlayer/ Runlayer provides a control plane for approved AI usage, an MCP gateway for agent-to-tool communications, and the Runlayer Agents runtime for agents created on Runlayer infrastructure. Why teams reach for it: Provides MCP governance, observability, and protections so regulated teams can adopt AI without exposing sensitive data. Worth knowing: Its primary function is access governance, scanning MCP server config, tool descriptions, agent skills. It supports shadow MCP discovery, prompt-injection detection, tool-poisoning detection. Specifications: - Primary Function: Access governance - Secondary Functions: Runtime guardrails, Posture management, Threat detection - Deployment Point: In request path - Scan Target: MCP server config, tool descriptions, agent skills - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Cursor, Claude Code, Codex, ChatGPT, VS Code, GitHub Copilot, Windsurf - Policy Enforcement: Blocking - Siem Integration: Yes --- ### Snyk - Vendor: Snyk - Category: Security & Governance - Build stage: Govern & protect - Deployment: SaaS - Licence: Not applicable - Website: https://snyk.io/ - Docs: https://docs.snyk.io/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--snyk/ An independent security layer that continuously validates AI-generated code, governs development agents, and secures AI-native applications—so organizations can move fast with AI without losing control. Why teams reach for it: Addresses machine-speed AI attacks, vulnerable AI-generated code, ungoverned development agents, and AI-application risk before production. Worth knowing: Its primary function is static scanning, scanning source code. Specifications: - Primary Function: Static scanning - Secondary Functions: Posture management, Runtime guardrails - Deployment Point: CLI or CI - Scan Target: source code - Siem Integration: Yes --- ### Ultra Security - Vendor: Ultra Security - Category: Security & Governance - Build stage: Govern & protect - Deployment: Self-hosted - Licence: Not applicable - Website: https://ultra.security/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--ultra-security/ Provides AI-native security infrastructure for MCP. Why teams reach for it: Pairs connector- and tool-level allow/deny rules with an LLM judge that scores each call over full session history. Worth knowing: Its primary function is runtime guardrails, scanning MCP server config, tool descriptions, network traffic. It supports shadow MCP discovery, prompt-injection detection, tool-poisoning detection. Specifications: - Primary Function: Runtime guardrails - Secondary Functions: Access governance, Threat detection, Static scanning, Data loss prevention, Posture management - Deployment Point: In request path - Scan Target: MCP server config, tool descriptions, network traffic - Shadow Mcp Discovery: Yes - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Not client-specific - Policy Enforcement: Both - Siem Integration: Yes --- ### Zenity - Vendor: Zenity - Category: Security & Governance - Build stage: Govern & protect - Deployment: Not disclosed - Licence: Not applicable - Website: https://zenity.io/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/security-governance--zenity/ Provides intent-aware AI-agent threat detection and response with visibility into execution paths, decisions, and actions. Why teams reach for it: Helps teams understand and stop harmful AI-agent decisions arising from external manipulation, internal exposure, or autonomous behavior. Worth knowing: It supports prompt-injection detection, tool-poisoning detection. Specifications: - Deployment Point: In request path - Prompt Injection Detection: Yes - Tool Poisoning Detection: Yes - Clients Covered: Amazon Bedrock AgentCore - Policy Enforcement: Both - Compliance Frameworks: OWASP, MITRE ATLAS --- ## Gateways & Proxies Layers that sit between a client and one or more MCP servers to route, aggregate, enforce policy, or federate. Many also terminate authentication; they are listed here when routing or aggregation is the primary function. ### agentgateway - Vendor: Linux Foundation - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Rust, Go, TypeScript - Website: https://agentgateway.dev/ - Docs: https://agentgateway.dev/docs/ - Source: https://github.com/agentgateway/agentgateway - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--agentgateway/ Open-source proxy for MCP and A2A with multiplexing, authorization, and OpenTelemetry. Why teams reach for it: It centralizes access to multiple MCP servers and applies authorization policies at the gateway. Worth knowing: Donated to the Linux Foundation by Solo.io, which offers a commercial distribution. Specifications: - Transports Client Facing: SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference, A2A, gRPC - Deploy Targets: single binary, Kubernetes - Self Hostable: Yes --- ### Bifrost - Vendor: maximhq - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Go, TypeScript, Python, JavaScript - Website: https://github.com/maximhq/bifrost - Docs: https://docs.getbifrost.ai - Source: https://github.com/maximhq/bifrost - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--bifrost/ Open-source LLM and MCP gateway with a unified API across providers, fallbacks and load balancing Why teams reach for it: Unifies access to 23+ AI providers behind one OpenAI-compatible API with automatic failover, load balancing, and semantic caching. Worth knowing: It speaks Streamable HTTP to clients, aggregates multiple MCP servers behind one endpoint, terminates authentication, can host stdio servers remotely. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: No - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference - Deploy Targets: Docker, Kubernetes - Self Hostable: Yes --- ### ContextForge - Vendor: IBM - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Python - Website: https://ibm.github.io/mcp-context-forge/latest/ - Source: https://github.com/IBM/mcp-context-forge - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--contextforge/ An open-source registry and proxy that federates MCP, A2A, and REST or gRPC APIs with centralized governance, discovery, and observability. Why teams reach for it: Presents a mixed estate of MCP servers and existing APIs to clients as one catalogue with shared governance. Worth knowing: It speaks stdio, SSE, Streamable HTTP to clients, aggregates multiple MCP servers behind one endpoint, terminates authentication, can host stdio servers remotely. Specifications: - Transports Client Facing: stdio, SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: A2A, REST, gRPC - Deploy Targets: Docker, Kubernetes, managed cloud - Self Hostable: Yes --- ### Docker MCP Gateway - Vendor: Docker - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: Go - Website: https://docs.docker.com/ai/mcp-catalog-and-toolkit/mcp-gateway/ - Source: https://github.com/docker/mcp-gateway - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--docker-mcp-gateway/ The MCP Gateway is Docker's open source solution for orchestrating Model Context Protocol (MCP) servers. It acts as a centralized proxy between clients and servers, managing configuration, credentials, and access control. Why teams reach for it: Runs third-party MCP servers in isolated Docker containers that, by default, do not receive the user's host environment and can access host paths only through validated mounts. Worth knowing: Pairs with the Docker MCP Catalog, listed under Registries & discovery. Specifications: - Transports Client Facing: stdio, SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference, A2A, REST, gRPC - Deploy Targets: Docker, Kubernetes, managed cloud - Self Hostable: Yes --- ### Lunar.dev MCPX - Vendor: Lunar.dev - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Hybrid - Licence: MIT - Languages: TypeScript - Website: https://www.lunar.dev/tutorials - Docs: https://docs.lunar.dev/mcpx/ - Source: https://github.com/TheLunarCompany/lunar - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--lunar-mcpx/ Aggregates multiple MCP servers behind a unified gateway with security, observability, and management. Why teams reach for it: Centralises which MCP tools are reachable and records how they are used. Worth knowing: It speaks Streamable HTTP to clients, aggregates multiple MCP servers behind one endpoint, terminates authentication, can host stdio servers remotely. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: Both - Deploy Targets: Docker - Self Hostable: Yes --- ### mcp-proxy - Vendor: sparfenyuk - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: Python - Website: https://github.com/sparfenyuk/mcp-proxy - Source: https://github.com/sparfenyuk/mcp-proxy - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--mcp-proxy/ Bridges Streamable HTTP and stdio MCP transports. Why teams reach for it: Switches between MCP server transports, bridging stdio and SSE/StreamableHTTP. Worth knowing: It speaks SSE, Streamable HTTP to clients, can host stdio servers remotely. Specifications: - Transports Client Facing: SSE, Streamable HTTP - Transports Server Facing: SSE, Streamable HTTP - Server Aggregation: No - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Delegates to external IdP - Tool Filtering: No - Agent Identity: No - Policy Enforcement: None - Telemetry Output: None - Also Proxies: LLM inference, A2A, REST, gRPC - Deploy Targets: Docker - Self Hostable: Yes --- ### mcpo - Vendor: Open WebUI - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: Python - Website: https://docs.openwebui.com/features/plugin/tools/openapi-servers/mcp - Source: https://github.com/open-webui/mcpo - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--mcpo/ Exposes an MCP server as an OpenAPI-compatible HTTP server, auto-generating interactive docs Why teams reach for it: Converts MCP server commands into standard RESTful OpenAPI HTTP servers so their tools work with OpenAPI-compatible agents and applications. Worth knowing: mcpo supports OAuth 2.1 authentication for MCP servers that require it. The implementation defaults to dynamic client registration, so most servers only need minimal configuration. Specifications: - Transports Client Facing: stdio, SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: No - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Delegates to external IdP - Tool Filtering: Yes - Agent Identity: No - Policy Enforcement: Built-in engine - Audit Logging: No - Telemetry Output: None - Deploy Targets: Docker - Self Hostable: Yes --- ### mcpproxy-go - Vendor: Smart MCP Proxy - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: Go, Swift - Website: https://mcpproxy.app/ - Docs: https://github.com/smart-mcp-proxy/mcpproxy-go - Source: https://github.com/smart-mcp-proxy/mcpproxy-go - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--mcpproxy-go/ Runs a local single-binary MCP proxy with a web UI, activity logging, and security quarantine. Why teams reach for it: Federate hundreds of MCP servers while bypassing Cursor's 40-tool limit and OpenAI's 128-function cap. Worth knowing: Linux Landlock enforcement via CI (ubuntu-latest), macOS graceful-degrade, scanner degradation under sandbox/none modes passes Specifications: - Transports Client Facing: stdio, SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Delegates to external IdP - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: Both - Also Proxies: REST - Deploy Targets: single binary, Kubernetes - Self Hostable: Yes --- ### MetaMCP - Vendor: Metatool - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: TypeScript - Website: https://metamcp.com/ - Docs: https://docs.metamcp.com/en - Source: https://github.com/metatool-ai/metamcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--metamcp/ Aggregates MCP servers into unified endpoints with middleware, authentication, and tool filtering Why teams reach for it: Reduces a long list of per-server client configurations to a small number of curated endpoints. Worth knowing: The product requires non-standard MCP protocol extensions for sampling feature Specifications: - Transports Client Facing: SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: No - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: None - Also Proxies: LLM inference, A2A, REST, gRPC - Deploy Targets: Docker, Kubernetes - Self Hostable: Yes --- ### MCP Gateway - Vendor: Microsoft - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: C#, TypeScript, Bicep, Python - Website: https://microsoft.github.io/mcp-gateway/ - Docs: https://github.com/microsoft/mcp-gateway - Source: https://github.com/microsoft/mcp-gateway - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--microsoft-mcp-gateway/ Reverse proxy and management layer for MCP servers, enabling scalable, session-aware stateful routing and lifecycle management of MCP servers in Kubernetes environments. Why teams reach for it: A reverse proxy and management layer for MCP servers that provides scalable, session-aware routing, authorization, and lifecycle management in Kubernetes. Worth knowing: It speaks Streamable HTTP to clients, aggregates multiple MCP servers behind one endpoint, can host stdio servers remotely. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Delegates to external IdP - Rbac: Yes - Tool Filtering: No - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference, A2A, REST, gRPC - Deploy Targets: Kubernetes, Docker - Self Hostable: Yes --- ### MintMCP - Vendor: MintMCP - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Hybrid - Licence: Not applicable - Website: https://www.mintmcp.com/ - Docs: https://www.mintmcp.com/docs/intro - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--mintmcp/ Enterprise gateway for the Model Context Protocol (MCP) providing observability, governance, and enablement for AI tools across ChatGPT, Claude, Cursor, and more. Why teams reach for it: Gives each agent a dedicated identity through the gateway, so agent tool calls are separable from human traffic in audit logs. Worth knowing: MintMCP uses a role-based access control model where roles are orthogonal to Virtual MCP access; custom roles cannot grant full admin access, and directory groups can map to roles for enterprise integration. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: stdio, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Delegates to external IdP - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: REST - Deploy Targets: Docker, managed cloud - Self Hostable: Yes --- ### Natoma - Vendor: Natoma - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://natoma.ai/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--natoma/ Gateway with a verified MCP server library, role-based profiles, a Cedar policy engine and identity-aware authorization Why teams reach for it: Brings unsanctioned MCP and agent use under one governed endpoint, pairing agentless discovery of shadow AI with Cedar attribute-based policy on every tool call. Worth knowing: It aggregates multiple MCP servers behind one endpoint. Specifications: - Server Aggregation: Yes - Rbac: Yes - Tool Filtering: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: Both - Also Proxies: MCP only - Deploy Targets: managed cloud - Self Hostable: Yes --- ### Obot MCP Gateway - Vendor: Obot AI - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: Go - Website: https://obot.ai/ - Docs: https://docs.obot.ai/ - Source: https://github.com/obot-platform/obot - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--obot-mcp-gateway/ The MCP Gateway is a reverse-proxy passthrough that sits between MCP clients and MCP servers. It authenticates users, ensures servers are deployed, and forwards requests without modifying the MCP protocol. Why teams reach for it: Gives an organisation one place to decide which MCP servers exist and who may reach them. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference - Deploy Targets: Kubernetes, Docker - Self Hostable: Yes --- ### Portkey MCP Gateway - Vendor: Portkey - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Hybrid - Licence: MIT - Languages: TypeScript, JavaScript - Website: https://portkey.ai/ - Docs: https://portkey.ai/docs/product/mcp-gateway - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--portkey-mcp-gateway/ The MCP Gateway is Portkey’s solution for managing access to MCP servers. It acts as a proxy between MCP Clients and MCP servers, handling authentication, access control, and logging. Why teams reach for it: Centralizes workspace-level rate limits for MCP tool calls without changing existing LLM rate-limit policies. Worth knowing: MCP Gateway handles authentication, access control, and logging for MCP servers; requires API keys or OAuth for authorization; supports multiple providers and models; includes observability features. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: Streamable HTTP - Server Aggregation: No - Hosts Stdio Servers Remotely: No - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference, A2A, REST, gRPC - Deploy Targets: Kubernetes - Self Hostable: Yes --- ### supergateway - Vendor: Supercorp - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: MIT - Languages: TypeScript - Website: https://github.com/supercorp-ai/supergateway - Source: https://github.com/supercorp-ai/supergateway - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--supergateway/ Runs MCP stdio servers over SSE, WebSockets or Streamable HTTP with one npx command Why teams reach for it: Exposes stdio-only MCP servers over SSE or WebSockets for remote access, debugging, and compatibility with clients that do not support stdio. Worth knowing: It speaks stdio, SSE, Streamable HTTP to clients, can host stdio servers remotely. Specifications: - Transports Client Facing: stdio, SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: No - Hosts Stdio Servers Remotely: Yes - Terminates Auth: No - Rbac: No - Tool Filtering: No - Agent Identity: No - Policy Enforcement: None - Audit Logging: No - Telemetry Output: None - Also Proxies: MCP only - Deploy Targets: Docker - Self Hostable: Yes --- ### ToolHive - Vendor: Stacklok - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Go, Python - Website: https://docs.stacklok.com/toolhive/ - Source: https://github.com/stacklok/toolhive - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--toolhive/ Deploys, runs, and manages MCP servers locally or in Kubernetes, with a gateway that aggregates multiple MCP servers into a unified interface. It also discovers, builds, distributes, installs, and manages Agent Skills for AI coding assistants. Why teams reach for it: ToolHive lets enterprises self-host MCP infrastructure so sensitive data need not be processed by SaaS providers. Worth knowing: ToolHive is a Kubernetes operator for MCP (Multi-Protocol Communication) servers. It requires Kubernetes clusters and provides SDKs for GitHub and other tools. The examples show both client and server components, and the status is active. Specifications: - Transports Client Facing: SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: No - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference, A2A, REST, gRPC - Deploy Targets: single binary, Docker, Kubernetes - Self Hostable: Yes --- ### Traefik Hub MCP Gateway - Vendor: Traefik Labs - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: Not applicable - Website: https://traefik.io/traefik-hub - Docs: https://doc.traefik.io/traefik-hub/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--traefik-hub-mcp-gateway/ MCP gateway in the Traefik Hub API platform with task-based access control Why teams reach for it: Separates MCP capability discovery from runtime authorization so visibility and invocation can be governed independently. Worth knowing: It speaks Streamable HTTP to clients, terminates authentication. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: Streamable HTTP - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference, REST, gRPC - Deploy Targets: Kubernetes - Self Hostable: Yes --- ### TrueFoundry AI Gateway - Vendor: TrueFoundry - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Hybrid - Licence: Not applicable - Website: https://www.truefoundry.com/ai-gateway - Docs: https://www.truefoundry.com/docs/ai-gateway/intro-to-llm-gateway - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--truefoundry/ Hosts stdio MCP servers behind the AI Gateway with centralized access, authentication, guardrails, and observability. Why teams reach for it: Routes both model calls and MCP tool calls through one gateway instead of running separate LLM and tool proxies. Worth knowing: It speaks SSE, Streamable HTTP to clients, aggregates multiple MCP servers behind one endpoint, terminates authentication, can host stdio servers remotely. Specifications: - Transports Client Facing: SSE, Streamable HTTP - Transports Server Facing: stdio, SSE, Streamable HTTP - Server Aggregation: Yes - Hosts Stdio Servers Remotely: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: OpenTelemetry - Also Proxies: LLM inference - Deploy Targets: managed cloud - Self Hostable: Yes --- ### WSO2 MCP Gateway - Vendor: WSO2 - Category: Gateways & Proxies - Build stage: Route & mediate - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Java - Website: https://wso2.com/api-platform/ - Docs: https://apim.docs.wso2.com/ - Source: https://github.com/wso2/product-apim - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/gateways-proxies--wso2-mcp-gateway/ Transforms existing APIs and OpenAPI definitions into managed MCP tools, and proxies existing MCP servers with governance, security, analytics, and lifecycle management. Why teams reach for it: It reaches MCP from the API-management side: the platform that already fronts an organisation's REST APIs auto-generates MCP servers from those APIs and then governs the resulting MCP traffic itself, acting as a resource server that validates JWT access tokens and serves protected-resource metadata at /.well-known/oauth-protected-resource, and allowing or denying individual tools, resources, prompts and JSON-RPC methods by scope and claim. Specifications: - Transports Client Facing: Streamable HTTP - Transports Server Facing: Streamable HTTP - Server Aggregation: Yes - Terminates Auth: Yes - Rbac: Yes - Tool Filtering: Yes - Agent Identity: Yes - Policy Enforcement: Built-in engine - Audit Logging: Yes - Telemetry Output: Both - Also Proxies: LLM inference - Deploy Targets: Docker, Kubernetes - Self Hostable: Yes --- ## Hosting & Deployment Platforms for running a remote MCP server in production, including the build-and-deploy pipelines that get code from a repository to a reachable endpoint. ### Alpic - Vendor: Alpic - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://alpic.ai/ - Docs: https://docs.alpic.ai/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--alpic/ Alpic provides the infrastructure layer that puts your product inside the AI conversation, and keeps it there at production scale. Why teams reach for it: Combines MCP server deployment, monitoring, analytics, and distribution in one cloud platform. Worth knowing: It deploys via Git push, dashboard, with custom domains. Specifications: - Deploy Method: Git push, dashboard - Auth Built In: Yes - Logs And Traces: Built-in - Custom Domain: Yes --- ### Blaxel - Vendor: Blaxel - Category: Hosting & Deployment - Build stage: Deploy - Deployment: Hybrid - Licence: Not applicable - Languages: Python, TypeScript - Website: https://blaxel.ai/ - Docs: https://docs.blaxel.ai/Overview - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--blaxel/ TypeScript-first SDK with Python at parity, providing Docker images, MCP servers, and storage built in for running agents. Why teams reach for it: Sandbox vendors were built for ephemeral execution, but Blaxel is built for agents that need to run, connect, and operate at scale. Worth knowing: It deploys via CLI, supporting Python, TypeScript, with custom domains. Specifications: - Deploy Method: CLI - Runtime Languages: Python, TypeScript - Auth Built In: Yes - Logs And Traces: Built-in - Custom Domain: Yes - Regions: Oregon, Virginia, London, and Frankfurt --- ### Cloudflare Agents - Vendor: Cloudflare - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: MIT - Languages: TypeScript, JavaScript - Website: https://developers.cloudflare.com/agents/ - Docs: https://developers.cloudflare.com/agents/tools/mcp/ - Source: https://github.com/cloudflare/agents - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--cloudflare-agents/ Provides persistent, stateful execution environments for agentic workloads, powered by Cloudflare Durable Objects, with per-agent state, storage, and lifecycle plus support for real-time communication, scheduling, AI model calls, MCP, and workflows. Why teams reach for it: Runs an internet-reachable MCP server with session persistence and authorization without operating servers. Worth knowing: The Agents SDK is open source under MIT; the platform it runs on is a commercial service. Specifications: - Deploy Method: Git push - Framework Lock In: Specific languages only - Runtime Languages: TypeScript, JavaScript - Auth Built In: Via integration - Session State Handling: Application responsibility - Logs And Traces: Built-in - Custom Domain: Yes - Byoc: No - Regions: Cloudflare's global network --- ### ContexaAI - Vendor: ContexaAI - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: Not applicable - Website: https://www.contexaai.com/ - Docs: https://docs.contexaai.com/github - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--contexaai/ ContexaAI is a developer platform for building, testing, debugging, deploying, and managing MCP servers. Why teams reach for it: It standardizes how agents connect to tools and data with less integration code. Worth knowing: It deploys via dashboard, supporting Any, and a bring-your-own-cloud option. Specifications: - Deploy Method: dashboard - Framework Lock In: Any framework - Runtime Languages: Any - Auth Built In: Yes - Logs And Traces: Built-in - Byoc: Yes --- ### Prefect Horizon - Vendor: Prefect - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://horizon.prefect.io/ - Docs: https://gofastmcp.com/deployment/running-server - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--fastmcp-cloud/ Provides an MCP gateway and governance layer for enterprise agents. Why teams reach for it: Gives a FastMCP server in a GitHub repo managed hosting at a production URL with built-in OAuth authentication. Worth knowing: It deploys via Git push, dashboard, supporting Python. Specifications: - Deploy Method: Git push, dashboard - Framework Lock In: Specific framework only - Runtime Languages: Python - Auth Built In: Yes - Logs And Traces: Built-in --- ### Golf.dev - Vendor: Golf - Category: Hosting & Deployment - Build stage: Deploy - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: Python - Website: https://www.golf.dev/ - Docs: https://docs.golf.dev/gateway/overview - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--golf-dev/ Golf is the single governance layer for all AI tool integrations. It deploys in your environment and sits between your AI tools and your systems—one control point for authentication, permissions, data inspection, and audit logging, regardless of which AI tool an employee uses. Why teams reach for it: It gives security teams visibility and control over MCP connections from third-party agents that bypass existing SIEM, DLP, and identity controls. Specifications: - Deploy Method: container image - Framework Lock In: Specific framework only - Runtime Languages: Python - Auth Built In: Yes - Session State Handling: Managed - Logs And Traces: Built-in - Byoc: Yes - Regions: Self-hosted rather than regional: Golf Gateway is available as a Docker image for Kubernetes and container deployments on the customer's own infrastructure, and the Control Plane "can be deployed wherever your organization requires" - either Golf Cloud, hosted and operated by Golf, or self-hosted. --- ### Manufact Cloud - Vendor: Manufact - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://manufact.com/ - Docs: https://docs.manufact.com/dashboard - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--manufact/ Manufact Cloud is an all-in-one platform for MCP servers: deploy from GitHub in minutes, watch every tool call with MCP-native analytics, and ship your server to any AI client, Cursor, Claude, ChatGPT, and more. Why teams reach for it: Hosts an MCP server and adds cross-client testing, analytics, and ChatGPT and Claude publishing checks. Specifications: - Deploy Method: Git push - Framework Lock In: Any framework - Runtime Languages: Any - Auth Built In: Yes - Logs And Traces: Built-in - Custom Domain: Yes - Regions: AUTO, US, EU or APAC compute region per server; all regions on the Startup plan --- ### mcp.run - Vendor: Dylibso - Category: Hosting & Deployment - Build stage: Deploy - Deployment: Self-hosted - Licence: BSD-3-Clause - Languages: Go, Rust, TypeScript, Python, Zig, C++, JavaScript - Website: https://turbomcp.ai/ - Docs: https://www.mcp.run/ - Source: https://github.com/dylibso/mcp.run-servlets - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--mcp-run/ Provided an official catalog of MCP servlets whose merged changes were automatically deployed to mcp.run users; the repository is now archived. Worth knowing: It supporting Go, Rust, TypeScript, JavaScript, Python, Zig, and a bring-your-own-cloud option. Specifications: - Framework Lock In: Specific framework only - Runtime Languages: Go, Rust, TypeScript, JavaScript, Python, Zig - Byoc: Yes --- ### Metorial - Vendor: Metorial - Category: Hosting & Deployment - Build stage: Deploy - Deployment: Self-hosted - Licence: Apache-2.0 - Languages: JavaScript, TypeScript, Python - Website: https://metorial.com/ - Docs: https://metorial.com/docs - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--metorial/ Connects AI systems and MCP clients to user-authorized tools, with portals, access control, and reviewable traces, sessions, tool calls, auth events, and failures. Why teams reach for it: You reach for Metorial when the hard part is not running an MCP server but controlling who may call it: it hosts the servers and handles the authentication and identity for each integration, puts every approved tool behind a single Magic MCP URL that people open with the company login, and - because the core is open source - can be self-hosted instead of taken as managed cloud. Specifications: - Deploy Method: dashboard - Framework Lock In: Any framework - Runtime Languages: Node.js - Auth Built In: Via integration - Session State Handling: Managed - Logs And Traces: Built-in - Byoc: Yes --- ### Naptha AI - Vendor: Naptha AI - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: Apache-2.0 - Languages: Python - Website: https://naptha.ai/ - Docs: https://www.naptha.ai/ - Source: https://github.com/NapthaAI/naptha-node - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--naptha-ai/ A framework and infrastructure for developing and running networked multi-agent systems at scale. Why teams reach for it: Runs multi-agent systems whose agents and modules sit on separate machines and still interact over the network, on a node you host yourself. Worth knowing: It supporting Python, TypeScript. Specifications: - Runtime Languages: Python, TypeScript - Session State Handling: Not applicable - Custom Domain: Not applicable --- ### Vercel - Vendor: Vercel - Category: Hosting & Deployment - Build stage: Deploy - Deployment: SaaS - Licence: Apache-2.0 - Languages: TypeScript, JavaScript - Website: https://vercel.com/docs/mcp - Source: https://github.com/vercel/mcp-handler - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--vercel/ Provides mcp-handler, a framework-agnostic HTTP adapter for hosting MCP servers in JavaScript and TypeScript applications. Why teams reach for it: Places an MCP server in the same deployment and domain as the application whose data it exposes. Worth knowing: The mcp-handler adapter is open source; the hosting platform is a commercial service. Specifications: - Deploy Method: Git push, CLI, dashboard - Framework Lock In: Specific languages only - Runtime Languages: Express, Flask, FastAPI, Hono, Nitro, Next.js, Python, Rust, Go, React, Vue, Angular, Svelte, Preact, Remix, Docusaurus, Hugo, Jekyll, Hexo, Polymer - Auth Built In: Via integration - Session State Handling: Application responsibility - Logs And Traces: Built-in - Custom Domain: Yes - Regions: Regions named in Vercel's regional pricing documentation: Cape Town, South Africa (cpt1); Cleveland, USA (cle1); Dubai, UAE (dxb1); Dublin, Ireland (dub1); Frankfurt, Germany (fra1); Hong Kong (hkg1); London, UK (lhr1); Montréal, Canada (yul1); Mumbai, India (bom1); Osaka, Japan (kix1); Paris, France (cdg1); Portland, USA (pdx1); San Francisco, USA (sfo1); São Paulo, Brazil (gru1); Seoul, South Korea (icn1); Singapore (sin1); Stockholm, Sweden (arn1); Sydney, Australia (syd1); Tokyo, Japan (hnd1); Washington D.C., USA (iad1). --- ### Zuplo - Vendor: Zuplo - Category: Hosting & Deployment - Build stage: Deploy - Deployment: Hybrid - Licence: MIT - Languages: TypeScript, JavaScript - Website: https://zuplo.com/ - Docs: https://zuplo.com/docs/mcp-server/tools - Source: https://github.com/zuplo/mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/hosting-deployment--zuplo/ Transforms selected OpenAPI routes into MCP tools, resources, and prompts through an MCP Server handler. Why teams reach for it: Publishes an existing API as an MCP server under the same policies already applied to its HTTP traffic. Worth knowing: It deploys via Git push, CLI, supporting Any, with custom domains, and a bring-your-own-cloud option. Specifications: - Deploy Method: Git push, CLI - Framework Lock In: Specific framework only - Runtime Languages: Any - Auth Built In: Yes - Session State Handling: Managed - Logs And Traces: Built-in - Custom Domain: Yes - Byoc: Yes - Regions: 300+ edge locations worldwide; single-tenant dedicated deployments available on Akamai, AWS, Azure, GCP and Equinix, or self-hosted on any Kubernetes. --- ## Inspectors & Debugging Tools for exercising an MCP server by hand: calling tools, reading responses, testing the authorization handshake, and checking conformance against the specification. ### Workers AI Playground - Vendor: Cloudflare - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://playground.ai.cloudflare.com/ - Docs: https://developers.cloudflare.com/agents/model-context-protocol/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--cloudflare-ai-playground/ Workers AI Playground is a browser-based tool for connecting to a remote MCP server by URL; once connected, the server's tools become available to an AI model in a built-in chat interface. Why teams reach for it: A hosted browser client where you paste a remote MCP server URL and its tools become available to a chat model. Specifications: - Interface: Web UI - Transports: Streamable HTTP - Oauth Flow Testing: Yes - Llm Playground: Yes - Install Method: Hosted --- ### FLUJO - Vendor: Mario Andreschak - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: Self-hosted - Licence: MIT - Languages: TypeScript, JavaScript, Python, Inno Setup - Website: https://github.com/mario-andreschak/FLUJO - Source: https://github.com/mario-andreschak/FLUJO - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--flujo/ FLUJO is an open-source, local-first application for building AI workflows by connecting models, MCP tools and reusable subflows on a drag-and-drop graph. Flows can be stepped through node by node in its built-in chat debugger, triggered on schedules or events, and served to other applications through OpenAI-compatible and MCP endpoints. Why teams reach for it: FLUJO is an MCP hub and inspector in one local-first app: it installs servers from the official MCP Registry, a GitHub repo or a local folder, lets you browse and call each server's tools, resources and prompts from its detail view with a one-click connection test, and re-exposes any configured server to other MCP clients such as Claude Desktop, Cursor or Cline over Streamable HTTP, so a server is set up and inspected once instead of per client. Specifications: - Interface: Web UI - Transports: stdio, SSE, Streamable HTTP - Tests Client Features: sampling, roots - Install Method: npx - Open Source: Yes --- ### mcp-cli - Vendor: IBM - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: Apache-2.0 - Languages: Python - Website: https://github.com/IBM/mcp-cli - Source: https://github.com/chrishayuk/mcp-cli - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcp-cli/ MCP CLI is a command-line client for Model Context Protocol servers, with a chat mode, a command-driven interactive shell, and a scriptable command mode for pipelines. It connects those servers' tools to LLMs from multiple providers and defaults to a local Ollama model that needs no API key. Why teams reach for it: Developers pick mcp-cli to drive MCP servers from a terminal or shell pipeline using local models without API keys. Specifications: - Interface: CLI, Web UI - Transports: stdio, SSE, Streamable HTTP - Tests Client Features: None - Conformance Testing: None - Install Method: uvx - Requires Account: No --- ### MCP Conformance Test Framework - Vendor: Model Context Protocol - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: MIT - Languages: TypeScript - Website: https://github.com/modelcontextprotocol/conformance - Source: https://github.com/modelcontextprotocol/conformance - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcp-conformance/ A framework for testing MCP (Model Context Protocol) client and server implementations against the specification. Why teams reach for it: An SDK can run exactly what a chosen spec revision requires, with every JSON-RPC message on the wire additionally validated against that revision's JSON schema, and a YAML baseline of known failures so the run exits non-zero only on a new regression or a stale baseline entry. Specifications: - Interface: CLI - Transports: Streamable HTTP - Oauth Flow Testing: Yes - Llm Playground: No - Tests Client Features: sampling, roots, elicitation - Conformance Testing: Full suite - Install Method: npx - Requires Account: No - Open Source: Yes --- ### MCP Inspector - Vendor: Model Context Protocol - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: Self-hosted - Languages: TypeScript - Website: https://github.com/modelcontextprotocol/inspector - Docs: https://modelcontextprotocol.io/docs/tools/inspector - Source: https://github.com/modelcontextprotocol/inspector - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcp-inspector/ MCP Inspector is a tool for connecting to an MCP server and exercising it by hand: browsing its resources, listing and running its prompts and tools, and watching the JSON-RPC, HTTP and stdio-stderr traffic of the session. It is published as the npm package `@modelcontextprotocol/inspector` and provides three clients - a web UI, a command-line client and a terminal UI - dispatched from one `mcp-inspector` binary. Why teams reach for it: One npm package gives web, CLI and terminal clients for the same MCP server, sharing OAuth tokens between them. Specifications: - Interface: Web UI, CLI - Transports: stdio, SSE, Streamable HTTP - Oauth Flow Testing: Yes - Tests Client Features: sampling, roots, elicitation - Conformance Testing: Partial - Install Method: npx --- ### mcp-remote - Vendor: Glen Maddern - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: MIT - Languages: TypeScript, JavaScript - Website: https://github.com/geelen/mcp-remote - Source: https://github.com/geelen/mcp-remote - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcp-remote/ A command-line client for connecting to remote MCP servers with OAuth authentication and support for multiple transport protocols including SSE and HTTP. Why teams reach for it: It lets an MCP client that only speaks local stdio talk to a remote MCP server, carrying the OAuth authorization flow those clients do not yet implement, and is meant to be dropped once the client supports remote authorized servers itself. Specifications: - Interface: CLI - Transports: stdio, SSE, Streamable HTTP - Oauth Flow Testing: Yes - Llm Playground: No - Tests Client Features: None - Conformance Testing: None - Install Method: npx - Requires Account: No - Open Source: Yes --- ### mcp-validator - Vendor: Janix.ai - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: AGPL-3.0-only - Languages: Python - Website: https://github.com/Janix-ai/mcp-validator - Source: https://github.com/Janix-ai/mcp-validator - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcp-validator/ Testing suite and reference implementation for validating MCP server implementations against the protocol specification. Why teams reach for it: It answers whether a server actually conforms to a named MCP revision - 2024-11-05, 2025-03-26 or 2025-06-18 - over STDIO or HTTP, checking the specification's MUST, SHOULD and MAY requirements and generating a compliance report, and it ships GitHub Actions templates so that check runs on every pull request rather than by hand. Specifications: - Interface: CLI - Transports: stdio, Streamable HTTP - Oauth Flow Testing: Yes - Conformance Testing: Full suite - Open Source: Yes --- ### MCPJam Inspector - Vendor: MCPJam - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: Hybrid - Licence: Apache-2.0 - Languages: TypeScript - Website: https://www.mcpjam.com/ - Docs: https://docs.mcpjam.com/ - Source: https://github.com/MCPJam/inspector - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcpjam-inspector/ Test, debug, and evaluate MCP servers across multiple AI clients. Chat with models to exercise tools, resources, and auth flows. Why teams reach for it: Tests one MCP server against 16 different client configurations and 170+ models, catching the ways ChatGPT, Claude and Cursor each read it differently, instead of debugging a single connection. Specifications: - Interface: Web UI, CLI, Desktop app - Transports: stdio, SSE, Streamable HTTP - Oauth Flow Testing: Yes - Llm Playground: Yes - Tests Client Features: elicitation - Conformance Testing: Full suite - Install Method: npx - Requires Account: Hosted features only - Open Source: Open core --- ### MCP Tools - Vendor: Fatih Kadir Akın - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: MIT - Languages: Go - Website: https://github.com/f/mcptools - Source: https://github.com/f/mcptools - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--mcptools/ A command-line interface for discovering, calling, and managing MCP server tools, resources, and prompts. Why teams reach for it: Drives any MCP server from the terminal, where output pipes straight into shell scripts, and can itself stand in as a mock, proxy or access-restricting guard server. Worth knowing: It provides a CLI interface over stdio, SSE, Streamable HTTP. Specifications: - Interface: CLI - Transports: stdio, SSE, Streamable HTTP - Oauth Flow Testing: No - Llm Playground: No - Tests Client Features: None - Conformance Testing: None - Install Method: Homebrew - Requires Account: No - Open Source: Yes --- ### Postman - Vendor: Postman - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://www.postman.com/ - Docs: https://learning.postman.com/docs/postman-ai-agent-builder/mcp-requests/overview/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--postman/ Postman is an API platform for designing, testing, documenting, and monitoring APIs, with Git and API gateway integrations across the API lifecycle. It also lets users create MCP requests to connect to, experiment with, and evaluate Model Context Protocol servers from within their collections. Why teams reach for it: Lets a team test and evaluate MCP servers using the same collections and workspace they already use for API testing. Specifications: - Interface: CLI, Desktop app, IDE extension - Transports: stdio, Streamable HTTP - Oauth Flow Testing: Yes - Llm Playground: Yes - Tests Client Features: sampling, elicitation --- ### Smithery CLI - Vendor: Arcade.dev - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: AGPL-3.0 - Languages: TypeScript, JavaScript - Website: https://github.com/smithery-ai/cli - Docs: https://smithery.ai/docs - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--smithery-cli/ Smithery CLI is a command-line tool that connects AI agents to MCP servers and skills. It allows users to search, add, and interact with MCP server tools directly from the terminal. Why teams reach for it: Smithery CLI keeps the whole loop in the terminal: it searches the Smithery registry, adds a server as a named connection, then lists, searches, inspects and calls that server's tools with smithery tool call, with --json for machine-readable output and --verbose for detailed logs when troubleshooting. Worth knowing: Smithery joined Arcade.dev; the CLI repository is now arcadeai-labs/smithery-cli. Specifications: - Interface: CLI - Transports: stdio, Streamable HTTP - Conformance Testing: None - Open Source: Yes --- ### @wong2/mcp-cli - Vendor: wong2 - Category: Inspectors & Debugging - Build stage: Test & debug - Deployment: CLI tool - Licence: GPL-3.0 - Languages: JavaScript - Website: https://github.com/wong2/mcp-cli - Source: https://github.com/wong2/mcp-cli - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/inspectors-debugging--wong2-mcp-cli/ mcp-cli is a command-line inspector for the Model Context Protocol that runs MCP servers from various sources or connects to an already-running server over Streamable HTTP or SSE. It lists and calls tools, reads resources and prompts, supports OAuth for SSE and Streamable HTTP servers, and offers a non-interactive mode for scripting. Why teams reach for it: Inspects an MCP server straight from a terminal with a single npx invocation, reusing the Claude Desktop config file format and adding a non-interactive mode that runs one tool, resource or prompt for scripting. Specifications: - Interface: CLI - Transports: stdio, SSE, Streamable HTTP - Oauth Flow Testing: Yes - Llm Playground: No - Tests Client Features: None - Conformance Testing: None - Install Method: npx - Requires Account: No - Open Source: Yes --- ## Registries & Discovery Catalogues that let a person or an agent find MCP servers. Includes the official registry, community directories, and container catalogues. ### Awesome MCP Servers - Vendor: punkpeye - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: Not disclosed - Licence: MIT - Languages: Not applicable - Website: https://github.com/punkpeye/awesome-mcp-servers - Source: https://github.com/punkpeye/awesome-mcp-servers - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--awesome-mcp-servers/ Awesome MCP Servers is a curated list of Model Context Protocol server implementations, published as the README of a GitHub repository and grouped into topic sections such as aggregators, browser automation and cloud platforms. Each entry carries markers for whether it is an official implementation, which language it is written in, which operating systems it targets, and whether it is a local service talking to software on the machine or a cloud service talking to a remote API; the same list is mirrored by a web directory at glama.ai/mcp/servers. Why teams reach for it: Awesome MCP Servers is a single curated index of MCP servers in which every entry is already tagged with its implementation language, its operating system, whether it is the official implementation, and whether it talks to locally installed software or to a remote API, so candidate servers can be shortlisted on those attributes and the same curation is browsable as a synced web directory rather than only as a repository README. Specifications: - Submission Model: Curated --- ### Docker MCP Catalog - Vendor: Docker - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://hub.docker.com/mcp - Docs: https://docs.docker.com/ai/mcp-catalog-and-toolkit/catalog/ - Source: https://github.com/docker/mcp-registry - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--docker-mcp-catalog/ The Docker MCP Catalog is a curated collection of verified MCP servers packaged as Docker images and distributed through Docker Hub. Each server runs as an isolated container, and Docker builds and signs the catalog's local servers. Why teams reach for it: Teams wanting MCP servers packaged as Docker images, avoiding the environment conflicts and setup complexity of running servers locally. Specifications: - Public Api: Yes - Submission Model: Curated - One Click Install Targets: Docker Desktop --- ### MCP.Directory - Vendor: MCP.Directory - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://fastmcp.me/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--fastmcp-me/ A directory of Model Context Protocol servers and agent skills that can be browsed, searched and compared. Each server page carries an Install button that produces one-click install configurations for Cursor, Claude Desktop, VS Code and Claude Code. Why teams reach for it: Puts MCP servers and agent skills in one searchable catalogue where each listing carries a one-click install config. Specifications: - Public Api: Yes - One Click Install Targets: Cursor, Claude Desktop, VS Code, Claude Code --- ### Glama - Vendor: Glama - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://glama.ai/mcp/servers - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--glama/ Glama is a web directory of Model Context Protocol servers and remote connectors that builds, runs and introspects each listed server in a sandbox and publishes the resulting schemas, behavioural findings and quality scores. It also deploys open-source servers from the registry onto its own infrastructure behind an MCP gateway that logs every JSON-RPC call, gates individual tools and manages OAuth credentials. Why teams reach for it: Every listed server is sandbox-built, introspected and scored, so its schemas and behaviour are checked before you install it. Specifications: - Public Api: Yes - Submission Model: Automated crawl - One Click Install Targets: None --- ### Kong MCP Registry - Vendor: Kong - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://konghq.com/products/kong-konnect - Docs: https://developer.konghq.com/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--kong-mcp-registry/ Kong MCP Registry is a centralized publishing and discovery endpoint for MCP servers within an organization, offered as part of Kong Konnect's Catalog. Each published server definition can list installable packages and hosted remote endpoints, and only clients authenticated to Konnect can query the registry. Why teams reach for it: Organisations already on Kong Konnect can register and discover MCP servers using the platform's existing authentication. Specifications: - Public Api: Yes - Submission Model: Open submission --- ### MCP.so - Vendor: MCP.so - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://mcp.so/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--mcp-so/ MCP.so is a community-driven directory of third-party MCP servers and clients, searchable by name, category, tag, runtime or integration. The same site also indexes remote servers, CLI tools, loops, agent skills and agents, and takes listings by submission, including a $39 paid submission that publishes immediately without review and carries a verified badge and featured placement. Why teams reach for it: Someone picks MCP.so over a servers-only registry because the same index also covers remote servers, clients, CLI tools, loops, agent skills and agents, and because a project author is buying reach: its submission page quotes 2.2M unique visitors and 6M pageviews over twelve months, and $39 buys immediate publication with a verified badge and a dofollow link. Specifications: - Submission Model: Open submission - One Click Install Targets: None --- ### MCPB - Vendor: Model Context Protocol - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: CLI tool - Licence: MIT - Languages: TypeScript, JavaScript - Website: https://github.com/modelcontextprotocol/mcpb - Source: https://github.com/modelcontextprotocol/mcpb - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--mcpb/ MCP Bundles (.mcpb) are zip archives containing a local MCP server and a manifest.json that describes the server and its capabilities, letting end users install a local MCP server with a single click. The repository provides the bundle specification, a CLI tool for creating bundles, and the code Claude for macOS and Windows uses to load and verify bundles; the project was renamed from DXT (Desktop Extensions). Why teams reach for it: MCPB turns a local MCP server into one installable file that Claude for macOS and Windows can add with a single click. Specifications: - One Click Install Targets: Claude for macOS and Windows --- ### mcpservers.org - Vendor: mcpservers.org - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://mcpservers.org/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--mcpservers-org/ mcpservers.org is a web directory of Model Context Protocol servers, browsable by category and filterable to featured or official entries. It also indexes remote MCP servers, agent skills, MCP clients and topic guides, and has a page for submitting new entries. Why teams reach for it: Covers MCP servers, remote endpoints, agent skills and clients in one categorised index, with topic guides for common use cases. Specifications: - Submission Model: Open submission --- ### mpak - Vendor: NimbleBrainInc - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: Hybrid - Licence: Apache-2.0 - Languages: TypeScript, Python - Website: https://mpak.dev/ - Source: https://github.com/NimbleBrainInc/mpak - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--mpak/ Open-source registry for MCP server bundles with cryptographic ownership verification and security scanning. Why teams reach for it: Unlike a metaregistry that only aggregates listings, mpak hosts the bundles themselves, runs 25 MCP-specific security controls across five domains on every publish and shows the resulting trust score and L1-L4 certification on each package; and because the whole stack is Apache 2.0, an organisation that will not point at mpak.dev can run its own instance with federation, policies and audit logging. Specifications: - Public Api: Yes - Submission Model: Verified ownership required --- ### MCP Registry - Vendor: Model Context Protocol - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: MIT - Languages: Go - Website: https://registry.modelcontextprotocol.io/ - Docs: https://registry.modelcontextprotocol.io/docs - Source: https://github.com/modelcontextprotocol/registry - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--official-mcp-registry/ MCP Registry is a hosted metadata repository for publicly available MCP servers, with a REST API that MCP clients and aggregators query to discover them. Server creators publish a single server.json entry giving the server's unique name, where to locate it and how to execute it, and the registry stores that metadata pointing at package registries such as npm, PyPI and Docker Hub rather than hosting the code itself. Why teams reach for it: It is the canonical record the rest of the ecosystem reads from - a server creator publishes once and MCP clients, aggregators and marketplaces all reference the same data - and the registry adds no ranking, curation or quality judgement of its own, leaving that to the subregistries that ETL from it. Specifications: - Public Api: Yes - Submission Model: Verified ownership required --- ### OpenTools - Vendor: OpenTools - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://opentools.com/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--opentools/ OpenTools is a hosted API and registry that lets developers call Model Context Protocol (MCP) tools through a single, OpenAI-compatible completions endpoint. It also provides a public registry for browsing available MCP servers, covering capabilities such as web search, location data, and web scraping. Why teams reach for it: OpenTools removes the need for separate API keys and billing per tool, and lets teams switch LLM providers without rewriting tool-calling code. --- ### PulseMCP - Vendor: PulseMCP - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://www.pulsemcp.com/ - Docs: https://www.pulsemcp.com/api - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--pulsemcp/ A directory of Model Context Protocol servers, clients and use cases, published alongside The Agentic Loop newsletter. PulseMCP also runs a Sub-Registry API that serves server.json metadata piped from the official MCP Registry together with its own enrichments, such as popularity estimates and a flag for whether the service owner maintains the server. Why teams reach for it: Surfaces which MCP servers are new or changed through daily updates and a versioned sync API, rather than a static list. Specifications: - Public Api: Yes - Submission Model: Curated - One Click Install Targets: None --- ### Smithery - Vendor: Arcade.dev - Category: Registries & Discovery - Build stage: Publish & discover - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://smithery.ai/ - Docs: https://smithery.ai/docs - Source: https://github.com/smithery-ai - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/registries-discovery--smithery/ Smithery is a registry where users find MCP servers and developers publish their own for distribution. Servers can also be hosted on Smithery infrastructure, and Smithery manages the OAuth flows and credential storage for connections to them. Why teams reach for it: Connects agents to a catalog of MCP servers while handling OAuth, credentials and session lifecycle instead of per-server setup. Specifications: - Public Api: Yes - Submission Model: Open submission - One Click Install Targets: Claude Desktop, Cursor --- ## Observability & Evaluation Tracing, logging, analytics, and evaluation for MCP servers and the agents that call them, in production rather than at development time. ### Arize Phoenix - Vendor: Arize AI - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Python, TypeScript, Jupyter Notebook, C, JavaScript, MDX, PLpgSQL - Website: https://phoenix.arize.com/ - Docs: https://arize.com/docs/phoenix - Source: https://github.com/Arize-ai/phoenix - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--arize-phoenix/ Arize Phoenix is an open-source observability platform for LLM applications, covering tracing, evaluation, prompt playgrounds, datasets and experiments. It can be self-hosted, and AI assistants such as Claude Code, Claude Desktop and Cursor can connect to a Phoenix instance over MCP to explore its projects, traces, prompts, datasets and experiments. Why teams reach for it: Teams pick Phoenix when LLM tracing and evaluation must run on open-source, self-hosted infrastructure with data never leaving their own network. Specifications: - Perspective: Agent developer - Mcp Aware: Both - Instrumentation Method: OpenTelemetry, client hooks - Traces: Yes - Evaluations: Yes - Datasets: Yes --- ### Braintrust - Vendor: Braintrust - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://www.braintrust.dev/ - Docs: https://www.braintrust.dev/docs - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--braintrust/ Braintrust logs production traces from LLM applications and agents and runs scored evaluations of their outputs against datasets. Prompts, scorers, experiments, and human review of logged data sit in the same interface. Why teams reach for it: Ties production traces to scored evaluations, so agent changes are measured in CI before shipping. Specifications: - Perspective: Agent developer - Mcp Aware: Generic LLM or agent - Instrumentation Method: server SDK, OpenTelemetry, client hooks - Traces: Yes - Session Replay: Yes - Evaluations: Yes - Datasets: Yes - Self Hostable: Enterprise plan only --- ### Helicone - Vendor: Helicone - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Apache-2.0 - Languages: TypeScript - Website: https://www.helicone.ai/ - Docs: https://docs.helicone.ai/ - Source: https://github.com/Helicone/helicone - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--helicone/ Helicone is an LLM observability platform that logs an application's requests to model providers and shows traces, sessions, cost and latency for them. It also runs an AI Gateway, an OpenAI-compatible API that reaches over 100 models with routing and automatic fallbacks. Why teams reach for it: Adding LLM observability is a one-line change - pointing an OpenAI client's baseURL at Helicone's AI Gateway - and the same open-source platform routes across 100+ models with automatic fallbacks and can be self-hosted from the project's docker-compose file. Specifications: - Perspective: Agent developer - Instrumentation Method: server SDK, proxy - Traces: Yes - Session Replay: Yes - Evaluations: No - Datasets: Yes - Self Hostable: Yes --- ### Ithena - Vendor: Ithena - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Library - Licence: Apache-2.0 - Languages: TypeScript - Website: https://ithena.one - Docs: https://github.com/ithena-one/mcp-governance-sdk/blob/main/docs/getting-started.md - Source: https://github.com/ithena-one/mcp-governance-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--ithena/ Ithena is a governance layer for MCP servers built on the TypeScript `@modelcontextprotocol/sdk`: its open-source `@ithena-one/mcp-governance` SDK wraps the base `Server` class in a `GovernedServer` that adds pluggable identity resolution, role-based access control, credential injection, auditing, request-scoped logging and trace-context propagation. A hosted Ithena Managed Platform providing backends for those interfaces is described as upcoming, with a waitlist. Why teams reach for it: The base MCP TypeScript SDK implements the protocol but leaves identity, authorization, secret access and audit trails to each server, so Ithena supplies them as one pluggable framework wrapping the existing `Server` class instead of every server reimplementing them. Specifications: - Instrumentation Method: server SDK - Traces: No --- ### Laminar - Vendor: Laminar - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Apache-2.0 - Languages: TypeScript, Rust - Website: https://www.lmnr.ai/ - Docs: https://docs.lmnr.ai/ - Source: https://github.com/lmnr-ai/lmnr - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--laminar/ Laminar is an OpenTelemetry-native observability platform for AI agents that traces every LLM call, tool call and sub-agent in a run. Alongside tracing it provides Signals, which read each run for behaviours described in plain English and cluster the matches, plus evaluations, datasets, dashboards, and CLI and MCP access that let a coding agent query traces with SQL. Why teams reach for it: Laminar surfaces agent failures nobody thought to instrument for - Signals read every run against behaviours described in plain English and cluster what matches - and exposes the same traces over SQL through a CLI and MCP server, so a coding agent can investigate a failure and replay the run while fixing it. Specifications: - Perspective: Agent developer - Instrumentation Method: server SDK, OpenTelemetry - Traces: Yes - Session Replay: Yes - Evaluations: Yes - Datasets: Yes - Self Hostable: Yes --- ### Langfuse - Vendor: Langfuse - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Apache-2.0 - Languages: TypeScript, JavaScript, Python, MDX, PLpgSQL - Website: https://langfuse.com/ - Docs: https://langfuse.com/docs - Source: https://github.com/langfuse/langfuse - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--langfuse/ Langfuse is an open-source AI engineering platform for tracing, evaluating and iterating on LLM applications and agents. Its hierarchical traces capture every LLM call, tool invocation and retrieval step, and it also provides prompt management, datasets, experiments and evaluators using LLM-as-a-judge, heuristic functions or human review. Why teams reach for it: Langfuse keeps tracing, prompt management, datasets, experiments and human annotation in one open-source platform a team can self-host, and its tracing is built on OpenTelemetry, so instrumentation is not tied to Langfuse's own SDK and the data can leave with the team. Specifications: - Perspective: Agent developer - Instrumentation Method: server SDK, OpenTelemetry - Traces: Yes - Evaluations: Yes - Datasets: Yes --- ### LangSmith - Vendor: LangChain - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://www.langchain.com/langsmith - Docs: https://docs.langchain.com/langsmith/home - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--langsmith/ LangSmith is an LLM application observability and evaluation platform. It traces applications built with any LLM framework, including via OpenTelemetry, and provides dashboards and alerting to monitor production performance. Why teams reach for it: Automatically clusters traces into recurring failure modes, and works with any LLM framework, not only LangChain. Specifications: - Perspective: Agent developer - Mcp Aware: Both - Instrumentation Method: server SDK, OpenTelemetry - Traces: Yes - Session Replay: Yes - Evaluations: Yes - Datasets: Yes - Self Hostable: Enterprise plan only --- ### AgentCat - Vendor: AgentCat - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: SaaS - Licence: MIT - Languages: TypeScript - Website: https://mcpcat.io/ - Docs: https://docs.agentcat.com/ - Source: https://github.com/agentcathq/agentcat-typescript-sdk - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--mcpcat/ AgentCat is an analytics and observability service for MCP servers: you add its SDK to your server and it records each agent session, every tool call with its request, response and error, and a goal inferred for the session. SDKs are published for TypeScript/JavaScript, Python and Go, the service runs in AgentCat's cloud or on your own infrastructure, and it can forward the enriched data to Datadog, Sentry or PostHog over OpenTelemetry. Why teams reach for it: Reach for it when you need the agent's side of an MCP session rather than server metrics: it captures why a tool was called and not only that it was, groups sessions into goals, and forwards the enriched data on to Datadog, Sentry or PostHog instead of replacing them. Specifications: - Perspective: Server operator - Instrumentation Method: server SDK - Traces: Yes - Session Replay: Yes --- ### Opik - Vendor: Comet - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Python, TypeScript - Website: https://www.comet.com/site/products/opik/ - Docs: https://www.comet.com/docs/opik/ - Source: https://github.com/comet-ml/opik - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--opik/ Opik is an Apache-2.0 licensed platform for tracing and evaluating LLM applications and AI agents, recording each step an agent takes from user interactions through context retrieval to tool calls. It supports datasets, experiments, code-based metrics, LLM-as-a-judge and online evaluation, and runs either on Comet's hosted Opik Cloud or self-hosted. Why teams reach for it: Teams wanting LLM tracing and evaluation inside their own infrastructure can self-host the whole platform without a commercial licence. Specifications: - Perspective: Agent developer - Instrumentation Method: server SDK, OpenTelemetry - Traces: Yes - Evaluations: Yes - Datasets: Yes --- ### Shinzo - Vendor: Shinzo Labs - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Sustainable Use License - Languages: TypeScript, Python - Website: https://github.com/shinzo-labs - Docs: https://docs.shinzo.ai/ - Source: https://github.com/shinzo-labs/shinzo - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--shinzo-labs/ Shinzo is an open-source observability platform for AI agents and MCP servers, used to ingest and analyse their performance data. Telemetry is OpenTelemetry-compliant, stored in Postgres, and viewed through an analytics dashboard. Why teams reach for it: It models MCP semantics rather than generic spans, so tool executions, cross-server traces and protocol-level metrics are visible in the way an MCP server operator needs them. Specifications: - Perspective: Both - Mcp Aware: Both - Instrumentation Method: server SDK, OpenTelemetry - Traces: Yes - Session Replay: Yes - Self Hostable: Yes --- ### Traceloop - Vendor: Traceloop - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Python - Website: https://www.traceloop.com/ - Docs: https://www.traceloop.com/docs - Source: https://github.com/traceloop/openllmetry - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--traceloop/ Traceloop uses OpenTelemetry to monitor and trace LLM applications. Instrumentation is either the OpenLLMetry SDK installed in the application or Traceloop Hub, a proxy that LLM calls are routed through. Why teams reach for it: Detect LLM quality regressions and production failures before users encounter them. Worth knowing: It observes from the agent developer perspective, runs evaluations. Specifications: - Perspective: Agent developer - Mcp Aware: Both - Instrumentation Method: server SDK, proxy - Traces: Yes - Evaluations: Yes - Datasets: Yes - Self Hostable: Yes --- ### Weave - Vendor: Weights & Biases - Category: Observability & Evaluation - Build stage: Observe & evaluate - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://wandb.ai/site/weave/ - Docs: https://weave-docs.wandb.ai/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/observability-evaluation--weave/ Tracing and evaluation for LLM applications including tool calls Why teams reach for it: Prevent production agent reliability collapse by learning from real-world failures that offline testing misses. Worth knowing: It observes from the agent developer perspective, supports session replay, runs evaluations. Specifications: - Perspective: Agent developer - Mcp Aware: Generic LLM or agent - Instrumentation Method: server SDK, OpenTelemetry - Traces: Yes - Session Replay: Yes - Evaluations: Yes - Datasets: Yes - Self Hostable: Yes --- ## Credential & Connection Management Products that hold the third-party credentials an MCP server needs on a user's behalf, and platforms that manage the connections to those upstream services. ### ACI.dev - Vendor: ACI.dev - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: Apache-2.0 - Languages: Python - Website: https://www.aci.dev/ - Docs: https://www.aci.dev/docs - Source: https://github.com/aipotheosis-labs/aci - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--aci-dev/ ACI.dev is an open-source tool-calling platform that connects AI agents to more than 600 pre-built application integrations, with multi-tenant OAuth flows and secrets management, natural-language permission boundaries and dynamic tool discovery built in. Agents reach those integrations either as direct function calls through its SDK or through its Model Context Protocol servers: an Apps MCP server that exposes the functions of specific apps and can carry several apps at once, and a Unified MCP server that exposes two meta-functions for discovering and executing any function on the platform. Why teams reach for it: ACI.dev covers both halves of putting a SaaS account behind an agent: it stores each end-user's OAuth tokens and API keys against a Linked Account Owner ID so agents act on behalf of a specific user, and its Unified MCP server hands the agent two meta-functions to search for and execute any function on the platform at call time rather than loading every tool definition into the context window. Specifications: - Credential Storage Location: Vendor cloud --- ### Arcade.dev - Vendor: Arcade.dev - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: Hybrid - Licence: MIT - Languages: Python, TypeScript - Website: https://www.arcade.dev/ - Docs: https://docs.arcade.dev/ - Source: https://github.com/ArcadeAI/arcade-mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--arcade-dev/ Arcade is a runtime that sits between AI agents and the systems they act on: it hosts MCP servers and tools, handles the OAuth handshake and token refresh with each upstream API so that neither the MCP client nor the LLM sees the token, and enforces per-action authorization against the organisation's own identity provider. It ships a catalog of 7,500+ prebuilt tools across 81 MCP servers, an open-source Python framework and CLI (arcade-mcp) for building and deploying your own, and MCP gateways that federate tools from several MCP servers into a single URL for the agents and apps that consume them. Why teams reach for it: Arcade takes user-level authorization out of the agent and makes it the runtime's job: a tool declares the OAuth scopes it needs, Arcade prompts the end user, stores and refreshes the token, and injects it per call so the MCP client and the LLM never see it, while every action is authorized at runtime against the organisation's existing identity provider at the intersection of what the agent and that user may each do. Specifications: - Credential Storage Location: Either --- ### Composio - Vendor: Composio - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: MIT - Languages: TypeScript, Python, JavaScript, Swift - Website: https://composio.dev/ - Docs: https://docs.composio.dev/ - Source: https://github.com/ComposioHQ/composio - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--composio/ Composio gives AI agents authenticated access to third-party app toolkits such as Slack, Gmail and GitHub through per-user sessions, and refreshes the underlying OAuth tokens itself. Each session is reachable from its TypeScript and Python SDKs and its CLI, or as a hosted MCP endpoint that any MCP client can connect to. Why teams reach for it: Composio supplies the connector layer an agent is missing - more than 1,000 integrations whose OAuth it runs end to end, scoped per connector - and exposes exactly the same catalog two ways from one account: as native tools inside your own agent through the Python or TypeScript SDK, or as a hosted MCP endpoint that an agent you already run, such as Claude Code, Codex or Cursor, can point at without a provider package. Specifications: - Exposes Mcp Endpoint: Both - Credential Storage Location: Vendor cloud - Free Tier: Yes --- ### Gumloop - Vendor: Gumloop - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://www.gumloop.com/ - Docs: https://docs.gumloop.com/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--gumloop/ Gumloop is an AI agent platform: users build AI agents that use their own tools and data, with no code required. The agents are automated with triggers, schedules and Gumloop's API, and they reach external services such as Gmail, Slack and Salesforce through connectors that hold the authentication. Why teams reach for it: Teams build agents on their own tools and data without writing code. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Either - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II, GDPR - Free Tier: No --- ### Klavis AI - Vendor: Klavis AI - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: Hybrid - Licence: Apache-2.0 - Languages: Python, Go, TypeScript, JavaScript, MDX - Website: https://www.klavis.ai/ - Docs: https://docs.klavis.ai/ - Source: https://github.com/Klavis-AI/klavis - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--klavis-ai/ Klavis AI is a platform that runs hosted MCP servers connecting AI agents to SaaS tools: one API call creates a server instance for a given end user and returns that server's URL along with an OAuth authorization URL, and the servers carry built-in OAuth and multi-tenancy authentication. The same servers are published as container images to run yourself, and the repository also ships MCP client bots for Discord, Slack, Web and WhatsApp. Why teams reach for it: Klavis runs the MCP server for each end user and owns the OAuth handshake with the upstream SaaS API, so an agent reaches one of 100+ prebuilt integrations through a single API call that returns a server URL and an authorization URL, instead of the integrating team standing up a server and token store per tool. Specifications: - Free Tier: Yes --- ### Make MCP - Vendor: Make - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://developers.make.com/mcp-server - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--make-mcp/ Make MCP Server lets AI systems such as large language models run Make scenarios and manage the contents of a Make account, including scenarios, connections, webhooks, data stores, teams and organizations. It turns active and on-demand scenarios into callable tools for MCP clients such as Claude and ChatGPT. Why teams reach for it: Turns scenarios already built in a Make account into MCP tools on a Make-hosted server. Specifications: - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Vendor cloud - Free Tier: Yes --- ### Merge - Vendor: Merge - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://www.merge.dev/ - Docs: https://docs.merge.dev/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--merge/ Unified API and agent handler across HRIS, ATS, CRM, accounting, file storage and ticketing Why teams reach for it: Connects a product and its AI agents to customers' third-party systems through one API instead of many separately built integrations. Worth knowing: It handles the upstream OAuth handshake, stores credentials per end user, with credentials held in vendor cloud, and handles token refresh automatically. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Vendor cloud - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II, ISO 27001, HIPAA, GDPR --- ### Nango - Vendor: Nango - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: Hybrid - Licence: Elastic-2.0 - Languages: TypeScript - Website: https://www.nango.dev/ - Docs: https://docs.nango.dev/ - Source: https://github.com/NangoHQ/nango - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--nango/ Nango is an open-source platform for building product integrations that connects applications and AI agents to more than 900 external APIs, handling OAuth flows, credential storage and token refresh for each end user's connection. Integration logic is written as TypeScript functions that run on Nango's production runtime and can be called from an app or exposed to agents through Nango's API, SDKs, or MCP server. Why teams reach for it: Handles per-user token storage and refresh for many upstream APIs in one place. Specifications: - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Either - Compliance Certifications: SOC 2 Type II, HIPAA, GDPR - Free Tier: Yes --- ### Paragon - Vendor: Paragon - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://www.useparagon.com/ - Docs: https://docs.useparagon.com/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--paragon/ Paragon is integration infrastructure that a company embeds into its own application, using Paragon's SDK and user authentication, so it can display and manage third-party integrations for its end users. Each end user who connects an integration is tracked individually in Paragon as a 'Connected User', and through the same platform the company can call third-party API endpoints and listen for third-party webhook events. Why teams reach for it: Teams give agents authenticated access to third-party app tools without hosting their own integration and MCP infrastructure. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Either - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II, HIPAA, GDPR --- ### Pipedream Connect - Vendor: Pipedream - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://pipedream.com/connect - Docs: https://pipedream.com/docs/connect/mcp/developers - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--pipedream-connect/ Pipedream Connect provides built-in user authentication for third-party APIs accessed via MCP. It removes the need for developers to build authorization flows or handle token storage and refresh when making authenticated requests on behalf of their end users. Why teams reach for it: Avoids building and maintaining separate OAuth integrations for each of the thousands of third-party APIs an agent might need to call. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Vendor cloud - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II, HIPAA, GDPR - Free Tier: Yes --- ### Unified.to - Vendor: Unified.to - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://unified.to/ - Docs: https://docs.unified.to/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--unified-to/ Unified.to provides a single unified API and MCP server for reading and writing customer data across dozens of SaaS categories such as CRM, ATS, HRIS, and accounting. Each request passes through live to the underlying service rather than being served from a cache or sync job. Why teams reach for it: Unified.to appeals to teams that need live customer data without storing it themselves, unlike sync-and-cache integration platforms. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II, HIPAA, GDPR - Free Tier: No --- ### Enterprise MCP - Vendor: Workato - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://www.workato.com/ - Docs: https://docs.workato.com/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--workato-enterprise-mcp/ Workato Enterprise MCP is a Workato feature for creating MCP servers from Workato project assets, API collections, or external servers, with token-based or OAuth2 single sign-on access control for MCP clients such as Claude, Cursor, and Windsurf. It also offers a verified user access mode for MCP tools that requires OAuth 2.0 authorization code grant authentication. Why teams reach for it: It centralizes MCP server discovery and governance across an organization, so teams reuse existing servers instead of building and maintaining duplicate integrations. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Vendor cloud - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II, HIPAA, GDPR, ISO 27001, PCI DSS --- ### Zapier MCP - Vendor: Zapier - Category: Credential & Connection Management - Build stage: Connect systems - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://zapier.com/mcp - Docs: https://docs.zapier.com/mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/credentials-connections--zapier-mcp/ Zapier MCP is a Model Context Protocol interface to Zapier that lets an AI client take actions in third-party apps - sending messages, creating tasks, updating records - through the app connections already present in the user's Zapier account. Most servers use dynamic tool discovery by default, so the agent searches for and enables the actions it needs rather than having them configured in advance. Why teams reach for it: Anyone who already connected apps in Zapier gets those connections to any AI client without terminal, config files, or code. Specifications: - Handles Upstream Oauth: Yes - Per End User Credentials: Yes - Exposes Mcp Endpoint: Hosted MCP server - Credential Storage Location: Vendor cloud - Token Refresh Handled: Yes - Compliance Certifications: SOC 2 Type II --- ## Clients Applications that consume MCP servers — the other end of every connection you build. Listed because which clients support which parts of the specification determines what a server can rely on. ### ChatGPT - Vendor: OpenAI - Category: Clients - Build stage: Connect a client - Deployment: SaaS - Licence: Not applicable - Languages: Not applicable - Website: https://chatgpt.com/ - Docs: https://developers.openai.com/apps-sdk/ - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--chatgpt/ ChatGPT is a hosted chat application that connects to remote MCP servers over streamable HTTP and exposes their tools to the model in a conversation. A server can also register a UI resource so an MCP Apps component renders inline in the chat. Why teams reach for it: A server published to the universal plugin directory reaches ChatGPT users and can render its own inline interface. Specifications: - Platform: Desktop app, Web - Transports: Streamable HTTP - Remote Server Support: Yes - Oauth Support: Yes - Tools: Yes - Resources: Yes - Config Scope: user or global - One Click Install: Gallery --- ### Claude Code - Vendor: Anthropic - Category: Clients - Build stage: Connect a client - Deployment: Hybrid - Licence: Not applicable - Languages: Not applicable - Website: https://claude.com/claude-code - Docs: https://docs.claude.com/en/docs/claude-code/mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--claude-code/ Claude Code is an AI coding agent from Anthropic, used from the terminal, as an extension for VS Code and JetBrains IDEs, or from a browser, mobile app, or desktop app. It reads codebases with agentic search, and can read issues, write code, run tests, and submit pull requests on GitHub and GitLab. Why teams reach for it: Exercises a server inside an agent that edits a real repository, with project-scoped config teammates can share. Specifications: - Platform: Desktop app, Web, CLI, IDE extension - Transports: stdio, SSE, Streamable HTTP - Remote Server Support: Yes - Oauth Support: Yes - Client Registration Mode: CIMD, DCR, manual - Tools: Yes - Resources: Yes - Prompts: Yes - Config Scope: user or global, workspace or project --- ### Claude Desktop - Vendor: Anthropic - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: Not applicable - Languages: Not applicable - Website: https://claude.ai/download - Docs: https://modelcontextprotocol.io/docs/develop/connect-local-servers - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--claude-desktop/ Claude Desktop is a downloadable desktop application for using Claude, available for macOS and Windows. It connects to local MCP servers configured through a JSON file, and to remote MCP servers by adding Custom Connectors. Why teams reach for it: Runs MCP servers as local programs on your own computer, configured in a JSON file, with tool actions requiring your approval. Specifications: - Platform: Desktop app - Transports: stdio - Remote Server Support: Yes - Tools: Yes - Resources: Yes - Prompts: Yes - Config Scope: user or global --- ### Cline - Vendor: Cline - Category: Clients - Build stage: Connect a client - Deployment: Not disclosed - Licence: Apache-2.0 - Languages: TypeScript, JavaScript, Rust, Python - Website: https://cline.bot/ - Docs: https://docs.cline.bot/mcp/mcp-overview - Source: https://github.com/cline/cline - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--cline/ Cline is an open source coding agent that runs as a VS Code extension, a JetBrains plugin, a terminal CLI, or an embeddable SDK. It connects to MCP servers over stdio, Streamable HTTP or SSE so their tools can be used alongside its built-in ones. Why teams reach for it: Cline is open source and locked to neither an editor nor a model vendor: one agent core ships as a VS Code extension, a JetBrains plugin, an npm-installed CLI and an embeddable Node SDK, and it drives Claude, GPT, Gemini, a local Ollama or LM Studio model, or any OpenAI-compatible endpoint on your own key or your own weights, with every file edit and terminal command held for approval unless you turn auto-approve on. Specifications: - Platform: IDE extension, CLI - Transports: stdio, SSE, Streamable HTTP - Client Registration Mode: DCR - Tools: Yes - Open Source: Yes --- ### Continue - Vendor: Continue - Category: Clients - Build stage: Connect a client - Deployment: Not disclosed - Licence: Apache-2.0 - Languages: TypeScript - Website: https://continue.dev/ - Docs: https://docs.continue.dev/ - Source: https://github.com/continuedev/continue - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--continue/ Continue is an AI coding assistant delivered as IDE extensions and as a terminal command-line agent (`cn`) with TUI and headless modes. It connects to Model Context Protocol servers, which let it reach external tools, systems and databases from its agent mode. Why teams reach for it: Continue is an open-source assistant whose model wiring is entirely the developer's: chat, edit, apply, autocomplete, embeddings and reranking are each assigned a model role in a config file, providers range from hosted APIs to a local Ollama model, and the documented offline path - sideload the .vsix, turn off anonymous telemetry, point the default model at a local provider - lets the whole assistant run air-gapped inside VS Code or JetBrains. Specifications: - Platform: CLI, IDE extension - Transports: stdio, SSE, Streamable HTTP - Oauth Support: Yes - Tools: Yes - Config Scope: user or global, workspace or project - Open Source: Yes --- ### Cursor - Vendor: Anysphere - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: Not applicable - Languages: Not applicable - Website: https://cursor.com/ - Docs: https://cursor.com/docs/context/mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--cursor/ Cursor is a coding agent used to understand a codebase, plan and build features, fix bugs, and review changes. It connects to Model Context Protocol (MCP) servers configured in project or global `mcp.json` files, or registered programmatically through its extension API. Why teams reach for it: Developers wanting an AI coding agent in a desktop app, with MCP servers installed and managed from a built-in marketplace. Specifications: - Platform: Desktop app, Web, CLI - Transports: stdio, SSE, Streamable HTTP - Remote Server Support: Yes - Oauth Support: Yes - Client Registration Mode: DCR, manual - Tools: Yes - Resources: Yes - Prompts: Yes - Config Scope: user or global, workspace or project - One Click Install: Gallery --- ### Gemini CLI - Vendor: Google - Category: Clients - Build stage: Connect a client - Deployment: CLI tool - Licence: Apache-2.0 - Languages: TypeScript, JavaScript, Python, C# - Website: https://github.com/google-gemini/gemini-cli - Docs: https://google-gemini.github.io/gemini-cli/ - Source: https://github.com/google-gemini/gemini-cli - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--gemini-cli/ Gemini CLI is an open-source AI agent that runs in a terminal and connects to Google's Gemini models. It ships built-in tools for file operations, shell commands, web fetching and Google Search grounding, and supports the Model Context Protocol for connecting external tools. Why teams reach for it: Signing in with a personal Google account gives Gemini CLI a free tier of 60 requests a minute and 1,000 a day against Gemini 3 models with a 1M-token context window and no API key to manage, and the agent it drives is Apache-2.0 open source with Google Search grounding, file, shell and web-fetch tools built in, MCP servers for anything else, and a non-interactive mode for running it in scripts. Specifications: - Platform: CLI - Transports: stdio, SSE, Streamable HTTP - Oauth Support: Yes - Client Registration Mode: DCR - Tools: Yes - Resources: Yes - Prompts: Yes - Config Scope: user or global, workspace or project - Open Source: Yes --- ### Goose - Vendor: Block - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: Apache-2.0 - Languages: Rust, TypeScript - Website: https://goose-docs.ai/ - Docs: https://goose-docs.ai/docs/getting-started/using-extensions/ - Source: https://github.com/block/goose - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--goose/ Open-source agent as desktop app and CLI, connecting to MCP servers as extensions Why teams reach for it: goose is a general-purpose agent rather than a coding-only one - research, writing, automation and data analysis as well as code - and it runs on your own machine as a native desktop app, a full CLI and an embeddable API from the same Rust codebase, against 15+ model providers including a local Ollama model or an existing Claude, ChatGPT or Gemini subscription through ACP, so neither the interface nor the model vendor is fixed by the tool. Specifications: - Platform: Desktop app, CLI - Transports: stdio, Streamable HTTP - Client Registration Mode: CIMD, DCR, manual - Tools: Yes - Config Scope: user or global - Open Source: Yes --- ### LibreChat - Vendor: LibreChat - Category: Clients - Build stage: Connect a client - Deployment: Self-hosted - Licence: MIT - Languages: TypeScript, JavaScript, Go, Python - Website: https://www.librechat.ai/ - Docs: https://www.librechat.ai/docs/features/mcp - Source: https://github.com/danny-avila/LibreChat - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--librechat/ LibreChat is a self-hosted, open-source AI chat platform that brings multiple AI model providers together in a single interface. Beyond chat it provides AI agents, Model Context Protocol (MCP) support, artifacts, a code interpreter, custom actions, conversation search, and multi-user authentication. Why teams reach for it: One self-hosted deployment gives a whole organisation a single chat interface over Anthropic, OpenAI, Azure, Bedrock, Google and any OpenAI-compatible endpoint, with multi-user OAuth2, LDAP or email sign-in and a browser admin panel for users, groups and roles, rather than each person holding an account with a vendor's own chat service. Specifications: - Platform: Web - Transports: stdio, SSE, Streamable HTTP - Oauth Support: Yes - Client Registration Mode: DCR, manual - Tools: Yes - Config Scope: user or global - Open Source: Yes --- ### Open WebUI - Vendor: Open WebUI - Category: Clients - Build stage: Connect a client - Deployment: Self-hosted - Licence: BSD-3-Clause - Languages: Python, Svelte, JavaScript, TypeScript - Website: https://openwebui.com/ - Docs: https://docs.openwebui.com/ - Source: https://github.com/open-webui/open-webui - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--open-webui/ Open WebUI is a self-hosted AI platform, used through the browser, that connects to models served by Ollama or any OpenAI-compatible API, whether local or cloud-hosted. It is extended with Python plugins - Filters, Actions, Pipes, Tools and Skills - and reaches external services through MCP, MCPO and OpenAPI tool servers. Why teams reach for it: Open WebUI puts the chat client itself on infrastructure you run - self-hosted, able to run entirely offline, and pointed at Ollama or any OpenAI-compatible API - and it keeps MCP centralised: only administrators can register MCP servers, and they scope each one to particular users or groups with access control. Specifications: - Platform: Web, Desktop app - Transports: Streamable HTTP - Oauth Support: Yes - Client Registration Mode: DCR, manual - Tools: Yes - Prompts: No - One Click Install: No --- ### Visual Studio Code - Vendor: Microsoft - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: MIT - Languages: TypeScript - Website: https://code.visualstudio.com/ - Docs: https://code.visualstudio.com/docs/copilot/customization/mcp-servers - Source: https://github.com/microsoft/vscode - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--visual-studio-code/ Visual Studio Code is an open source code editor whose chat agent connects to MCP servers declared in an mcp.json file in a workspace or in the user profile. It supports the stdio, Streamable HTTP and legacy SSE transports, and consumes server tools, prompts, resources, elicitation, sampling, roots and MCP Apps. Why teams reach for it: MCP servers are installed from a gallery inside the editor and configured either in a checked-in .vscode/mcp.json that a team shares through source control or in a user profile that follows the developer across every workspace. Specifications: - Platform: Desktop app - Transports: stdio, SSE, Streamable HTTP - Remote Server Support: Yes - Oauth Support: Yes - Client Registration Mode: DCR, manual - Tools: Yes - Resources: Yes - Prompts: Yes - Config Scope: user or global, workspace or project, dev container - One Click Install: Both - Open Source: Open core --- ### Warp - Vendor: Warp - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: Not applicable - Languages: Not applicable - Website: https://www.warp.dev/ - Docs: https://docs.warp.dev/knowledge-and-collaboration/mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--warp/ Warp is an open source terminal for AI-assisted software development that runs coding agents alongside the shell on macOS, Linux and Windows. Its local agents connect to MCP servers added in the app or read from ~/.warp/.mcp.json and .warp/.mcp.json files, over a launched command, Streamable HTTP or SSE. Why teams reach for it: Puts the MCP-connected agent inside the terminal itself, reusing server definitions already configured for Claude Code or Codex. Specifications: - Platform: Desktop app, CLI - Transports: stdio, SSE, Streamable HTTP - Remote Server Support: Yes - Oauth Support: Yes - Tools: Yes - Resources: Yes - Config Scope: user or global, workspace or project - One Click Install: Gallery - Open Source: Yes --- ### Windsurf - Vendor: Cognition - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: Not applicable - Languages: Not applicable - Website: https://windsurf.com/ - Docs: https://docs.windsurf.com/windsurf/cascade/mcp - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--windsurf/ Windsurf is a desktop IDE for Mac, Windows, and Linux whose built-in Cascade agent acts as an MCP client, connecting to servers over stdio, Streamable HTTP, and SSE listed in ~/.codeium/windsurf/mcp_config.json. The vendor has rebranded the editor to Devin Desktop, and its Linux packages now install devin-desktop with windsurf kept as a transitional package. Why teams reach for it: Teams that want an editor whose agent installs MCP servers from a marketplace admins can allowlist or replace with their own registry. Specifications: - Platform: Desktop app - Transports: stdio, SSE, Streamable HTTP - Remote Server Support: Yes - Oauth Support: Yes - Tools: Yes - Resources: Yes - Prompts: Yes - Config Scope: user or global - One Click Install: Both --- ### Zed - Vendor: Zed Industries - Category: Clients - Build stage: Connect a client - Deployment: Desktop app - Licence: GPL-3.0 - Languages: Rust, Python, Inno Setup, Tree-sitter Query, WGSL, Rich Text Format, Metal, HLSL, JavaScript, TypeScript, Julia, Objective-C, RenderScript - Website: https://zed.dev/ - Docs: https://zed.dev/docs/ai/mcp - Source: https://github.com/zed-industries/zed - Last verified: 21 Aug 2026 - Page: https://mcptoolchain.ai/vendor/clients--zed/ Zed is a code editor for macOS, Linux and Windows. It uses the Model Context Protocol to connect to MCP servers, and currently supports their tools and prompts. Why teams reach for it: Zed is a code editor that is itself the MCP client: servers are installed from Zed's extension list or added by hand as local or remote entries under Settings → AI → MCP Servers - a remote server with no Authorization header triggers the standard MCP OAuth flow - and the servers configured there can also be forwarded over ACP to external agents. Specifications: - Platform: Desktop app - Transports: stdio, Streamable HTTP - Client Registration Mode: CIMD, DCR, manual - Tools: Yes - Resources: No - Prompts: Yes - Config Scope: user or global, workspace or project - One Click Install: Gallery - Open Source: Yes ---