Start free

Add the Cognee MCP server to Claude Code

Config last verified Nov 10, 2026

The exact config to run Cognee in Claude Code — paste it in, restart, and the tools load.

Prerequisites

  • Claude Code installed.
  • LLM_API_KEY — API key for the LLM provider Cognee uses to extract the graph and answer queries (OpenAI by default).

Setup

1. Open ~/.claude.json

2. Add this configuration

Add to ~/.claude.json

~/.claude.json
json
{
  "mcpServers": {
    "cognee": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "TRANSPORT_MODE=stdio",
        "-e",
        "LLM_API_KEY",
        "cognee/cognee-mcp:main"
      ],
      "env": {
        "LLM_API_KEY": "<LLM_API_KEY>"
      }
    }
  }
}
Or via CLI
bash
claude mcp add cognee -- docker run -i --rm -e TRANSPORT_MODE=stdio -e LLM_API_KEY cognee/cognee-mcp:main

3. Restart Claude Code and confirm the Cognee tools load.

Gotchas

Claude Code stores MCP servers under an "mcpServers" object in ~/.claude.json. It also reads a project-scoped .mcp.json and supports adding servers from the command line with "claude mcp add". Each entry must declare either a "type" or a "command"; Claude Code refuses an entry that has neither. Remote servers connect natively over Streamable HTTP with OAuth.

← Back to the Cognee MCP server
MCP Server DirectoryGlen home