This library provides seamless integration of Model Context Protocol (MCP) servers, enabling agentic frameworks to access over 650 specialized tools. It targets developers building AI agents who need to leverage external data and functionalities, offering a unified interface to a vast ecosystem of MCP-powered resources.

How It Works

MCPAdapt acts as a bridge, translating MCP server tool definitions (often in JSON Schema format) into the specific tool formats required by various agentic frameworks. It supports both local command execution and remote Server-Sent Events (SSE) connections to MCP servers, abstracting away the underlying communication protocols. This approach allows agents to discover and invoke MCP tools as if they were native functions.

Quick Start & Requirements

• Installation:
  • With Smolagents: uv add smolagents[mcp]
  • With other frameworks (e.g., Langchain): uv add mcpadapt[langchain] or pip install mcpadapt[langchain]
• Dependencies: Python 3.x. Specific framework integrations may have additional requirements.
• Resources: Requires an MCP server to be running or executable.
• Documentation: https://github.com/grll/mcpadapt

Highlighted Details

• Supports integration with Smolagents, Langchain, CrewAI, and Google GenAI.
• Enables seamless use of over 650 existing MCP servers.
• Provides support for both local command execution and remote SSE MCP servers.
• Includes an adapter pattern for easy extension to new agentic frameworks.

Maintenance & Community

Core maintainers include @grll, with contributions from @murawakimitsuhiro, @joejoe2, @tisDDM, and @sysradium.

Licensing & Compatibility

The repository does not explicitly state a license in the provided README. Users should verify licensing for commercial use or integration into closed-source projects.

Limitations & Caveats

The README includes a security warning advising caution when connecting to MCP servers, especially over SSE, and emphasizes verifying the source and security of any server before connecting. Support for Pydantic-AI and Swarm is listed as planned but not yet implemented.