@swarmv/mcp-server

MCP server for SwarmV: gives an AI agent persistent vector memory, limited to what the signed-in user is allowed to see and edit.

Tools:

• read_from_memory returns the closest records (id, distance, metadata) from the containers the user can view.
• write_to_memory stores a note in a container the user can edit (put the text in metadata.text).

Setup

Add it to your MCP client's configuration (Claude Desktop, Cursor and others):

{
  "mcpServers": {
    "swarmv": {
      "command": "npx",
      "args": ["-y", "@swarmv/mcp-server"],
      "env": {
        "SWARMV_TOKEN_URL": "https://your-backend.example.com/swarmv-token",
        "SWARMV_TOKEN_URL_AUTH": "<the signed-in user's session credential>"
      }
    }
  }
}

Credentials

Pick one:

| Variables | Use | | --- | --- | | SWARMV_TOKEN_URL (+ optional SWARMV_TOKEN_URL_AUTH) | Recommended. Your backend returns { "token": "..." }, a short-lived user token it minted with your API key. | | SWARMV_USER_TOKEN | A user token you already have (expires). | | SWARMV_API_KEY + SWARMV_USER_ID=user:<id> | Local development only. Never ship your API key to end users. |

Optional: SWARMV_DIM (vector dimensions, default 768), SWARMV_BASE_URL.

Full documentation: https://swarmv.dev/docs

License

MIT