LangChain MCP is a Model Context Protocol (MCP) server that provides semantic search across the entire LangChain ecosystem. Build AI applications faster with instant access to documentation and source code for LangChain, LangGraph, LangSmith, and DeepAgents.

• Semantic Search - Natural language queries across all LangChain ecosystem docs
• Source Code Search - Find code examples in Python and JavaScript repositories
• MCP Protocol - Works seamlessly with Claude Code, Claude Desktop, Cursor, and any MCP-compatible client
• Production Ready - Scalable API with authentication and usage tracking
• Fast & Accurate - Powered by ChromaDB and OpenRouter embeddings

# Install globally
npm install -g langchain-mcp

# Login with Google
langchain-mcp login

# Add to Claude Code
claude mcp add langchain-mcp -- npx langchain-mcp

Add the following configuration to your client's config file:

Claude Desktop

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Cursor

• macOS/Linux: ~/.cursor/mcp.json
• Windows: %USERPROFILE%\.cursor\mcp.json

{
  "mcpServers": {
    "langchain-mcp": {
      "command": "npx",
      "args": ["langchain-mcp"]
    }
  }
}

langchain-mcp login      # Login via Google OAuth
langchain-mcp status     # Check usage and remaining credits
langchain-mcp logout     # Logout and clear credentials

• Free tier available for new users
• Donation bonus for supporters

langchain-MCP/
├── packages/
│   ├── ingest/              # Python - Data ingestion (uv)
│   ├── api/                 # TypeScript - API server (Express)
│   ├── mcp-server/          # TypeScript - MCP client (npm package)
│   └── mcp-server-local/    # TypeScript - Local MCP server (dev)
├── config/
│   └── settings.json        # Shared configuration
└── deploy.sh                # Deployment script

1. Ingest Documentation & Source Code

cd packages/ingest
uv sync
uv run ingest --list     # List available repositories
uv run ingest docs       # Ingest documentation only
uv run ingest            # Ingest all (docs + code)

2. Run API Server

cd packages/api
npm install
npm run dev              # Development server on port 3000

3. Test Local MCP Server

cd packages/mcp-server-local
npm install
npm run dev

All settings in config/settings.json:

{
  "embedding": {
    "provider": "openrouter",
    "model": "qwen/qwen3-embedding-8b"
  },
  "chromadb": {
    "path": "./data/chroma"
  },
  "chunking": {
    "docs": { "chunk_size": 2000, "chunk_overlap": 200 },
    "code": { "chunk_size": 4000, "chunk_overlap": 200 }
  },
  "repos": [
    {
      "name": "langchain",
      "url": "https://github.com/langchain-ai/langchain",
      "type": "code",
      "languages": ["python", "javascript"]
    }
  ]
}

• sentence-transformer (local)
• openai
• cohere
• google
• ollama
• openrouter (default)

See ChromaDB Integrations for more options.

The project includes automated deployment scripts for VPS hosting:

# Manual deployment
./deploy.sh

# GitHub Actions (production branch)
git push origin main:production

Deployment includes:

• Code synchronization via rsync
• Automatic npm installation and build
• PM2 process management
• Nginx static file serving
• Environment variable management

• Semantic search across docs and code
• Google OAuth authentication
• Usage tracking and credits system
• MCP registry registration
• Claude Code, Desktop, and Cursor support
• Rate limiting (per user / per IP)
• Additional embedding model options
• Local mode (no API key required)
• Browser extension for quick searches
• VSCode extension integration

Forking and contributions are welcome!

• Website: langchain-mcp.xyz
• Issues: GitHub Issues
• Donate: Support development at Ko-fi

MIT License - see the LICENSE file for details.