Use MCP with AI Tools

Use MCP when you want an AI tool such as Claude Desktop, Claude Code, or another MCP-compatible client to store and retrieve data on Autonomi through structured tools instead of direct HTTP requests. This setup still uses antd under the hood, but the MCP server handles the bridge between your AI client and the daemon.

Prerequisites

• Python 3.10+

• An MCP-compatible client such as Claude Desktop or Claude Code

• antd built and running on your machine (see Start the Local Daemon)

Steps

1. Start the local daemon

The MCP server talks to antd, so start there first:

./target/release/antd

Run that command from the ant-sdk/antd build directory, or use antd if the binary is already on your PATH.

If you have not built antd yet, follow Build with the SDKs and Start the Local Daemon first.

2. Install the MCP server

From the ant-sdk repo root:

pip install "antd[rest]"
pip install -e antd-mcp/

3. Configure your AI client

For Claude Desktop, add antd-mcp to claude_desktop_config.json:

{
  "mcpServers": {
    "antd-autonomi": {
      "command": "antd-mcp",
      "env": {
        "ANTD_BASE_URL": "http://127.0.0.1:8082"
      }
    }
  }
}

Adjust ANTD_BASE_URL if your daemon runs on a different host or port.

4. Continue with the full MCP setup

Use the full setup guide next. It covers daemon discovery, SSE mode, and overrides such as ANTD_BASE_URL.

What happened

Your AI client talks to antd-mcp, and the MCP server calls antd on your behalf.

Next steps

• Use the Autonomi MCP Server

• MCP Server Reference

• Start the Local Daemon