MCP To LangChain Tools Conversion Utility / TypeScript

What is MCP To LangChain Tools Conversion Utility / TypeScript?

A lightweight TypeScript library that converts MCP (Model Context Protocol) server tools into LangChain-compatible StructuredTool objects. It allows developers to integrate any MCP server—local or remote—into LangChain agents and chains with minimal configuration.

How to use MCP To LangChain Tools Conversion Utility / TypeScript?

Install via npm i @h1deya/langchain-mcp-tools, then call convertMcpToLangchainTools() with an McpServersConfig object that mirrors the mcpServers structure from Claude for Desktop. The function returns an array of LangChain tools and a cleanup callback. Use the tools with LangChain models and always invoke cleanup() when done.

Key features of MCP To LangChain Tools Conversion Utility / TypeScript

• Converts multiple MCP servers into LangChain tools in parallel
• LLM provider-specific schema transformations (Gemini, OpenAI, etc.)
• Supports local stdio and remote HTTP, SSE, and WebSocket MCP servers
• Automatic transport selection and explicit configuration
• Stderr redirection and working directory control for local servers
• Concurrent cleanup with error handling

Use cases of MCP To LangChain Tools Conversion Utility / TypeScript

• Integrate filesystem, fetch, and search MCP tools into a LangChain agent
• Combine public and private MCP servers (Brave Search, GitHub, Notion) in one application
• Quickly experiment with custom MCP servers in a LangChain prototype
• Access the growing ecosystem of 2000+ MCP servers from LangChain

FAQ from MCP To LangChain Tools Conversion Utility / TypeScript

What are the prerequisites?

Node.js 18+. Optional: npm 7+ for Node.js-based local MCP servers, and uv (uvx) for Python-based local MCP servers.

Does it support remote MCP servers?

Yes. It supports Streamable HTTP, SSE, and WebSocket transports. Transport is auto-detected or can be set explicitly via a type or transport field.

What are the known limitations?

Only text results of tool calls are supported—image or other content types are filtered out. Only the MCP Tools feature is implemented; Resources, Prompts, and Sampling are not supported.

How does the library handle schema compatibility?

When the llmProvider option is specified, the library performs provider-specific JSON schema transformations to prevent compatibility issues (e.g., for Google Gemini or OpenAI). Set it if you encounter schema-related warnings or errors.

How does cleanup work?

The cleanup function closes all MCP server connections concurrently and logs any failures. It continues cleanup of remaining servers even if some fail. Always call it in a finally block after using the tools.