Cognee

Provides a Cognee memory engine with knowledge-graph construction and search capabilities via MCP.

typescript

GitHub Stars

typescript

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "mcp-mirror-topoteretes_cognee-mcp-server": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/your/cognee-mcp-server",
        "run",
        "mcpcognee"
      ],
      "env": {
        "ENV": "local",
        "DB_NAME": "cognee_db",
        "DB_PROVIDER": "sqlite",
        "LLM_API_KEY": "YOUR_API_KEY",
        "VECTOR_DB_PROVIDER": "lancedb",
        "TOKENIZERS_PARALLELISM": "false",
        "GRAPH_DATABASE_PROVIDER": "networkx"
      }
    }
  }
}

You run Cognee as an MCP server to empower your memory engine with structured querying and knowledge graph capabilities. This server exposes tools you can invoke from MCP clients, and it can be started locally using a command-line runtime. It also defines environment variables to control behavior, such as API keys and provider settings.

How to use

You will start the Cognee MCP server locally and connect an MCP client to it. The server provides a Cognify_and_search tool that builds a knowledge graph from input text, then retrieves relevant edges using a search query. To use it, start the MCP server in your environment, then issue requests from your client to create or query knowledge graphs, and the server will return the retrieved graph edges as results.

How to install

Prerequisites you need before starting the server: a compatible runtime environment (Node.js is commonly used for MCP servers with UV-based runners) and access to the command runtime suggested by the server configuration.

Prepare the environment variables shown in the configuration snippet. 2) Install any required runtimes or package managers if you plan to run the server locally. 3) Use the provided startup command to run the MCP server.

Configuration and runtime details

The server is designed to run locally using a command-line runtime and a set of environment variables. The configuration example uses a runtime runner to execute the MCP server module named mcpcognee in a directory you specify. It also sets environmental controls like the API key for the LLM, graph database provider, vector database provider, and database settings.

{
  "mcpcognee": {
    "command": "uv",
    "args": [
      "--directory",
      "/path/to/your/cognee-mcp-server",
      "run",
      "mcpcognee"
    ],
    "env": {
      "ENV": "local",
      "TOKENIZERS_PARALLELISM": "false",
      "LLM_API_KEY": “your llm api key”,
      "GRAPH_DATABASE_PROVIDER": “networkx”,
      "VECTOR_DB_PROVIDER": "lancedb",
      "DB_PROVIDER": "sqlite",
      "DB_NAME": “cognee_db”
    }
  }
}

Tools

The server exposes a Cognify_and_search tool that builds a knowledge graph from input text and performs a search over that graph. It accepts inputs to construct the graph and a query to retrieve relevant edges from the knowledge graph.

Available tools

Cognify_and_search

Builds a knowledge graph from input text and performs a search in the graph using a provided query. Outputs the retrieved edges of the knowledge graph.

Built by

VeilStrat

AI signals for GTM teams