Home / MCP / Documentation MCP Server
Provides access to documentation data for LangChain, LlamaIndex, and OpenAI via a searchable MCP tool.
Configuration
View docs{
"mcpServers": {
"documentation": {
"command": "uv",
"args": [
"--directory",
"/ABSOLUTE/PATH/TO/YOUR/documentation",
"run",
"main.py"
],
"env": {
"SERPER_API_KEY": "your_serper_api_key_here"
}
}
}
}You can integrate a Documentation MCP Server to let Claude search and extract information from LangChain, LlamaIndex, and OpenAI documentation directly within conversations. This server provides a focused, context-rich way to retrieve relevant docs and feed them into your chat workflows.
You connect to the Documentation MCP Server from your MCP client and start a local stdio server that runs the search and extraction logic. Once connected, you can ask Claude to look up information such as how vector stores are documented in LangChain or details about embedding models, and the tool will return concise documentation excerpts you can review in the chat.
Prerequisites: You need Python 3.11 or higher and the uv package manager. You also require a Serper API key for Google site-restricted searches.
# Install the uv package manager
curl -LsSf https://astral.sh/uv/install.sh | sh
# Clone and set up the project
git clone <repository-url> documentation
cd documentation
# Create and activate virtual environment
uv venv
# On Windows:
.venv\Scripts\activate
# On macOS/Linux:
source .venv/bin/activate
# Install dependencies
uv pip install -e .
# Create .env with your Serper API key
echo "SERPER_API_KEY=your_serper_api_key_here" > .envConfigure the client to run the MCP server locally. You will use a stdio-based setup where the startup command must be exactly as shown, including the directory path to your project and the script to run.
{
"mcpServers": {
"documentation": {
"command": "uv",
"args": [
"--directory",
"/ABSOLUTE/PATH/TO/YOUR/documentation",
"run",
"main.py"
]
}
}
}If Claude cannot locate the server, verify the directory path in your configuration is correct and that the virtual environment is activated before starting the server.
If searches return no results, double-check your Serper API key and network connectivity.
If you encounter timeouts, the server may be experiencing connectivity issues or rate limits from the search provider.
A tool to search for relevant documentation excerpts across LangChain, LlamaIndex, and OpenAI, returning concise excerpts based on your query, library, and max character limit.