agi

The AGI MCP Server provides persistent memory capabilities for AI systems by interfacing with the AGI Memory database. It supports episodic, semantic, procedural, and strategic memories, plus vector-based similarity search, memory clustering, identity persistence, worldview tracking, and graph-based memory relationships. Tools exposed by the server allow you to create, retrieve, search, and organize memories, as well as inspect system health and current thematic clusters. You can connect to the server from clients like Claude Desktop using a configuration payload that specifies the command to run the MCP server, the path to the server entrypoint, and the necessary environment variables for PostgreSQL and environment mode. The server is designed to work with the AGI Memory backend and relies on a PostgreSQL database with pgvector and Apache AGE graph extensions for rich memory representations.

Prerequisites:

• Node.js 18+
• Docker and Docker Compose (for AGI Memory DB setup)
• Git

1. Install AGI Memory database (external dependency)

• Clone the AGI Memory repo and set up the database as described in its README
• Start the memory database (e.g., via Docker Compose) and ensure PostgreSQL, pgvector, and Apache AGE are available

2. Install and run the MCP Server

• Clone this MCP server repository
• Install dependencies

  git clone https://github.com/cognitivecomputations/agi-mcp-server.git
  cd agi-mcp-server
  npm install
  

• Prepare environment variables

  cp .env.example .env
  # Edit .env with your actual database credentials and settings
  

• Start the MCP server

  npm start
  

3. Optional: Run directly from GitHub using npx

• Example Claude Desktop config (using npx)

  {
    "mcpServers": {
      "agi-memory": {
        "command": "npx",
        "args": [
          "-y",
          "github:cognitivecomputations/agi-mcp-server"
        ],
        "env": {
          "POSTGRES_HOST": "localhost",
          "POSTGRES_PORT": "5432",
          "POSTGRES_DB": "agi_db",
          "POSTGRES_USER": "agi_user",
          "POSTGRES_PASSWORD": "agi_password",
          "NODE_ENV": "development"
        }
      }
    }
  }
  

Notes:

• Ensure the memory database is reachable with the credentials you provide in the env vars.
• If you encounter path or ENOENT issues, consider using full paths to node/npx as shown in the README example.

Tips and common issues:

• Environment variables: POSTGRES_HOST, POSTGRES_PORT, POSTGRES_DB, POSTGRES_USER, POSTGRES_PASSWORD control database access. NODE_ENV can be development or production.
• When using GUI clients like Claude Desktop, ensure system-wide PATH includes Node.js binaries or use absolute paths for node/npx as described in the guide.
• If the server fails to initialize, check the AGI Memory database status (e.g., docker compose ps for the memory DB) and verify that the PostgreSQL extension pgvector and Apache AGE are active.
• Memory tools available include orientation (get_memory_health, get_active_themes, get_identity_core, get_worldview), search (search_memories_similarity, search_memories_text, get_memory_clusters, activate_cluster, get_memory), and creation (create_memory, create_memory_cluster).
• For testing, you can run a manual sample like: POSTGRES_HOST=localhost POSTGRES_PORT=5432 POSTGRES_DB=agi_db POSTGRES_USER=agi_user POSTGRES_PASSWORD=agi_password NODE_ENV=development node mcp.js