search-product-contentLABSCO SUMMARY
Your documentation, answerable from wherever the question comes up.
This Python server registers one tool: a retrieval over the knowledge base in your Inkeep project, under a name and description you choose. It returns the passages themselves, each with its title and source URL, so an assistant can quote your documentation and point to the page instead of paraphrasing from memory.
What it is
A Python MCP server that puts your Inkeep knowledge base behind one retrieval tool, named by INKEEP_MCP_TOOL_NAME (search-product-content in the README example), that any MCP client can call. Each call goes to the Inkeep API with your key, so the passages come from the knowledge base in your Inkeep project rather than from the model's memory.
What you get
- A retrieval tool that returns passages from your knowledge base for a question, under the name and description you configure
- Passages returned as citation documents, each with its title, source URL and record type, so an answer can link back to the page it came from
- A tool name and description you set yourself in INKEEP_MCP_TOOL_NAME and INKEEP_MCP_TOOL_DESCRIPTION, so a client with several servers connected can tell this one apart
- A local stdio server run from your own clone under uv, with the retrieval model picked in INKEEP_API_MODEL (inkeep-rag in the README example)
Requirements
An Inkeep account and project, with an API key created under Projects → Integrations → Create Integration → API. uv, and a clone of the repository that the client launches with uv --directory <YOUR_INKEEP_MCP_SERVER_ABSOLUTE_PATH> run -m inkeep_mcp_server. Three settings have no default and must be in the env block: INKEEP_API_BASE_URL (https://api.inkeep.com/v1), INKEEP_API_KEY and INKEEP_API_MODEL (inkeep-rag).
Setup effort
Build from source — git clone https://github.com/inkeep/mcp-server-python.git
USE CASES
01
Getting an answer without opening the docs site
Ask in the client and the tool answers from your knowledge base rather than from the model's recollection.
02
Finding the passage rather than the summary
The search tool returns the relevant sections so you can read them yourself.
03
Making sure the assistant asks this server
Give the tool a name and description for your product, so an assistant with several servers connected sends product questions here.
04
Linking an answer to its page
Each passage comes back with its title and source URL, so the reply can point to the documentation page it came from.
1 tool
Listed in project docs. Your client may show a different list.
BEFORE YOU INSTALL
What to know first
Tools you get
1 tool named in the project's documentation.
Requires
- uv on PATH
- INKEEP_API_KEY in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
You clone this one yourself.
Server setup
TERMINAL
git clone https://github.com/inkeep/mcp-server-python.git
cd mcp-server-python
uv venv
uv pip install -r pyproject.tomlMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"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. The query should be framed as a conversational question about Inkeep."
}
}
}
}Create INKEEP_API_KEY in the Inkeep Dashboard (Projects → your project → Integrations → Create Integration → API) and put it in the env block above, beside INKEEP_API_BASE_URL and INKEEP_API_MODEL; the server does not start without all three.
<YOUR_INKEEP_MCP_SERVER_ABSOLUTE_PATH> is the full path of your mcp-server-python clone. If the client cannot find uv, put its full path in command; which uv on macOS and Linux or where uv on Windows prints it.
