mcp-server-python

What is mcp-server-python?

mcp-server-python is an MCP server that provides retrieval-augmented generation (RAG) over your documentation and product content using an Inkeep account. It's designed for developers who want to query their own product docs through the Model Context Protocol.

How to use mcp-server-python?

Clone the repository, set up a Python virtual environment with uv, and install dependencies. Obtain an API key from the Inkeep Dashboard, then add the server configuration to your claude_desktop_config.json file, specifying the absolute project path, API key, and base URL.

Key features of mcp-server-python

• RAG powered by your docs and product content.
• Uses Inkeep’s API for retrieval.
• Single tool: search-product-content.
• Requires an Inkeep account and API key.
• Runs locally via uv Python project manager.
• Integrates with Claude Desktop via MCP.

Use cases of mcp-server-python

• Accessing product documentation through conversational queries in Claude Desktop.
• Providing context-aware answers based on your proprietary knowledge base.
• Enabling AI assistants to retrieve specific help articles or feature guides.
• Reducing manual search time by asking natural language questions about your product.

FAQ from mcp-server-python

What does the mcp-server-python do?

It serves as a bridge between your documentation (managed by Inkeep) and an MCP client, allowing AI assistants to retrieve relevant product content using the search-product-content tool.

What are the prerequisites to use this server?

You need an Inkeep account (to manage and serve your RAG), the uv Python project manager, and Python. An API key must be generated from the Inkeep Dashboard.

How do I configure it with Claude Desktop?

Add a JSON entry to claude_desktop_config.json under mcpServers, specifying the uv command, the project directory, and environment variables including INKEEP_API_BASE_URL, INKEEP_API_KEY, and INKEEP_MCP_TOOL_NAME.

Where is my data stored and processed?

Your documentation is stored and indexed by Inkeep; the server only sends queries and retrieves results via Inkeep’s API. No local document storage is used.

What tool does the server expose and how should queries be framed?

The server exposes a single tool named search-product-content. Queries should be framed as conversational questions about the product content.