Inkeep MCP Server for Product Documentation

Overview

Inkeep MCP Server is a lightweight Python implementation of a Model Context Protocol (MCP) server that connects a local MCP client to Inkeep’s Retrieval-Augmented Generation (RAG) service. It runs locally and forwards incoming MCP tool calls to your Inkeep project using an API key, enabling conversational agents (like Claude Desktop or other MCP-capable clients) to query product documentation and other product content dynamically.

This server is useful when you want to power interactive, searchable product documentation, assist onboarding, or augment support workflows without shipping your full documentation into the model. It handles authentication and request forwarding, letting your MCP client present a single tool that retrieves context-rich passages from your Inkeep project.

Features

• Simple MCP server that bridges local MCP clients to the Inkeep API
• Uses Inkeep RAG model for context-aware retrieval from your product docs
• Runs locally using uv (a Python project runner) and standard virtual environments
• Configurable via environment variables (API key, model, tool name/description)
• Ready to plug into MCP clients such as Claude Desktop via mcpServers configuration

Installation / Configuration

Prerequisites

• An Inkeep account and a project configured in the Inkeep Dashboard
• uv installed (https://github.com/astral-sh/uv)
• Python 3.8+

Clone and prepare the project locally:

git clone https://github.com/inkeep/mcp-server-python.git
cd mcp-server-python

# create virtual environment and install dependencies using uv
uv venv
uv pip install -r pyproject.toml

Obtain an Inkeep API key

1. Sign in to the Inkeep Dashboard (portal.inkeep.com).
2. Open your Project → Integrations → Create Integration → API.
3. Create the integration and copy the generated API key.

Environment variables Set the following environment variables before launching the server. You can export them in your shell or provide them via your MCP client launcher.

Example (Bash):

export INKEEP_API_KEY="sk-xxxx..."
export INKEEP_API_BASE_URL="https://api.inkeep.com/v1"
export INKEEP_API_MODEL="inkeep-rag"
export INKEEP_MCP_TOOL_NAME="search-product-content"
export INKEEP_MCP_TOOL_DESCRIPTION="Retrieves product documentation about our product."

Running the server Start the MCP server with uv:

uv run -m inkeep_mcp_server

If uv is not on your PATH, use the full path reported by which uv (macOS/Linux) or where uv (Windows).

Available Tools / Resources

The MCP server exposes a single tool (configurable via env vars) that accepts a conversational query and returns relevant content from your Inkeep project. Typical properties:

• Tool name: INKEEP_MCP_TOOL_NAME (e.g., search-product-content)
• Tool description: INKEEP_MCP_TOOL_DESCRIPTION
• Backend: Inkeep RAG model specified by INKEEP_API_MODEL
• Authentication: INKEEP_API_KEY sent to Inkeep API

Source code and issues: https://github.com/inkeep/mcp-server-python

Integrating with an MCP Client (example: Claude Desktop)

Add the server definition to your client config (replace placeholders):

{
  "mcpServers": {
    "inkeep-mcp-server": {
      "command": "uv",
      "args": [
        "--directory",
        "<YOUR_INKEEP_MCP_SERVER_ABSOLUTE_PATH>",
        "run",
        "-m",
        "inkeep_mcp_server"
      ],
      "env": {
        "INKEEP_API_BASE_URL": "https://api.inkeep.com/v1",
        "INKEEP_API_KEY": "<YOUR_INKEEP_API_KEY>",
        "INKEEP_API_MODEL": "inkeep-rag",
        "INKEEP_MCP_TOOL_NAME": "search-product-content",
        "INKEEP_MCP_TOOL_DESCRIPTION": "Retrieves product documentation about Inkeep."
      }
    }
  }
}

Notes:

• Replace <YOUR_INKEEP_MCP_SERVER_ABSOLUTE_PATH> with the repository absolute path.
• If uv is not in PATH, use its full path in the command field.

Use Cases

• Searchable product documentation
  • Allow conversational agents to answer technical questions by retrieving passages from docs, troubleshooting guides, and FAQs.
• Faster onboarding
  • New hires or customers can ask natural-language questions and get concise, sourced snippets from your documentation.
• Support assist and ticket triage
  • Surface relevant KB articles or configuration steps to accelerate first-response times.
• Embedded knowledge in chat workflows
  • Combine RAG answers from Inkeep with your assistant’s reasoning to provide accurate, context-linked responses.

Tips & Troubleshooting

• Ensure INKEEP_API_KEY has correct project scope in the Inkeep Dashboard.
• If the MCP client cannot start the server, confirm the uv command path and working directory are correct.
• For production, avoid exposing API keys in plaintext; use a secrets manager or environment-secured service.
• Inspect server logs to confirm requests are proxied to Inkeep and to debug model or network errors.

For full source, issues, and contribution guidelines, see the GitHub repository: https://github.com/inkeep/mcp-server-python