Zep Cloud MCP Server

An MCP (Model Context Protocol) server that wraps the Zep Cloud API, giving AI assistants long-term memory via Zep's knowledge graph.

Tools

| Tool | Description | |------|-------------| | zep_search_memory | Search a user's knowledge graph for facts | | zep_add_note | Add a fact/note to a user's knowledge graph | | zep_get_context | Get memory context for a thread (for prompt injection) | | zep_add_messages | Store messages in a thread for memory extraction | | zep_create_user | Create or ensure a Zep user exists | | zep_create_thread | Create a new conversation thread for a user |

Setup

npm install

Get your API key from Zep Cloud.

Usage

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "zep": {
      "command": "node",
      "args": ["/absolute/path/to/zep-mcp/server.mjs"],
      "env": {
        "ZEP_API_KEY": "your-api-key"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "zep": {
      "command": "node",
      "args": ["/absolute/path/to/zep-mcp/server.mjs"],
      "env": {
        "ZEP_API_KEY": "your-api-key"
      }
    }
  }
}

OpenClaw / mcporter

{
  "mcpServers": {
    "zep": {
      "command": "node",
      "args": ["./path/to/zep-mcp/server.mjs"],
      "env": {
        "ZEP_API_KEY": "your-api-key"
      }
    }
  }
}

Standalone

ZEP_API_KEY=your-api-key node server.mjs

The server communicates over stdio using JSON-RPC (MCP protocol).

How It Works

Zep automatically builds a knowledge graph from conversations. The typical workflow:

1. Create a user → zep_create_user
2. Create a thread → zep_create_thread
3. Store messages → zep_add_messages (Zep extracts facts automatically)
4. Query memories → zep_search_memory or zep_get_context
5. Add direct notes → zep_add_note (for facts not from conversation)

License

MIT