cognee-mcp-server

What is cognee-mcp-server?

An MCP server for cognee, an AI memory engine. It provides a tool that builds a knowledge graph from input text and performs search within that graph, returning retrieved edges.

How to use cognee-mcp-server?

Configure it in your Claude Desktop client (claude_desktop_config.json) using uvx. Set the required environment variables (e.g., LLM_API_KEY, GRAPH_DATABASE_PROVIDER, VECTOR_DB_PROVIDER, DB_PROVIDER, DB_NAME). Invoke the Cognify_and_search tool with the text and search query as inputs.

Key features of cognee-mcp-server

• Builds a knowledge graph from input text
• Performs search in the constructed graph
• Returns retrieved edges of the knowledge graph
• Supports optional custom Pydantic graph models

Use cases of cognee-mcp-server

• Ingesting unstructured text and answering queries over it
• Building a searchable memory layer for AI applications

FAQ from cognee-mcp-server

What does the Cognify_and_search tool do?

It constructs a knowledge graph from the provided text and then performs a search using the given query, returning the matching edges of the graph.

Does cognee-mcp-server support custom graph models?

Yes. You can optionally provide a graph_model_file (filename of a custom Pydantic graph model implementation) and graph_model_name (the class name) to customize the graph structure.

What environment configuration is required?

You must set LLM_API_KEY, GRAPH_DATABASE_PROVIDER, VECTOR_DB_PROVIDER, DB_PROVIDER, and DB_NAME. Example values: networkx, lancedb, sqlite.

What runtime dependencies does it have?

It requires Python and uv (for running via uvx). The optional custom graph model feature depends on a Pydantic implementation.