Gemini Docs

You can run a lightweight, remote HTTP MCP server that indexes Gemini API documentation and exposes search and retrieval capabilities. It starts on startup by ingesting Gemini docs, builds a fast local search index, and serves MCP endpoints for clients to query documentation content efficiently.

How to use

Connect your MCP client to the server over HTTP at the MCP endpoint to search and retrieve Gemini API docs. Use the search capability to locate relevant pages, fetch full page content, or obtain information about the current Gemini models. You can also run in local stdio mode during development for direct MCP interactions without listening on a network port.

How to install

Prerequisites: you need Python and/or a runtime to install and run the MCP server, and optionally Docker for containerized deployment. You will also need a command-capable environment to run or build the server.

Option 1: Use uvx to run the server directly without explicit installation.

uvx --from git+https://github.com/philschmid/gemini-api-docs-mcp gemini-docs-mcp

Option 2: Install directly from GitHub using Python’s package manager.

pip install git+https://github.com/philschmid/gemini-api-docs-mcp.git

Option 3: Manual installation (for development) to clone, install, and clean up after setup.

git clone https://github.com/philschmid/gemini-api-docs-mcp.git
cd gemini-api-docs-mcp
pip install -e .
cd ..
rm -rf gemini-api-docs-mcp

Running and configuring the server

Run the server remotely over HTTP or in local stdio mode. The HTTP server listens on a configurable port, defaulting to 8080, and exposes the MCP endpoint at /mcp. You can run it with a local command or deploy it to a cloud service.

To run in HTTP mode and expose the MCP endpoint on a specific port, set the port environment variable and start the server.

# Set port (optional, defaults to 8080)
export PORT=8080

# Run the server
gemini-docs-mcp

The MCP endpoint will be available at http://localhost:8080/mcp (or the port you configured). If you want to expose this to the internet or a cloud service, configure your load balancer or cloud runtime to route /mcp to the server.

## Configuration and environment

The server stores its local SQLite database with a full-text search index. You can override the default database path by setting GEMINI\_DOCS\_DB\_PATH. The server supports a default path in containers and a user’s home directory locally.

Default database paths:

• In containers: /tmp/gemini-api-docs/database.db
• Locally: ~/.mcp/gemini-api-docs/database.db

You can override with the environment variable GEMINI_DOCS_DB_PATH.

Using with an MCP client

For remote HTTP servers, configure your MCP client to connect via HTTP at the MCP URL path. Include the endpoint that serves the MCP protocol at /mcp in your client configuration.

{
  "mcpServers": {
    "gemini-docs": {
      "url": "https://<your-service-url>/mcp"
    }
  }
}

Security and notes

If you deploy to a public cloud, ensure your service is protected and only exposes the MCP endpoint to authorized clients. Use standard cloud security practices such as enabling HTTPS, restricting access, and monitoring usage.

Available tools

search_documentation

Performs a full-text search on Gemini documentation for given queries (up to 3).

get_capability_page

Fetches a list of capabilities or the content for a specific capability.

get_current_model

Retrieves documentation for the current Gemini models.