> ## Documentation Index
> Fetch the complete documentation index at: https://docs.magic-resume.cn/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect AI coding tools (MCP)

> Advanced — read and edit your Magic Resume resumes right inside Claude Code, Cursor, and similar tools.

If you use AI coding tools like **Claude Code** or **Cursor**, you can plug Magic Resume in so the AI can **read your resumes and safely edit them on your instructions** — using the same data layer as the web app.

<img src="https://mintcdn.com/magic-resume-web/sfC90NORLtOisICN/images/mcp-dataflow.png?fit=max&auto=format&n=sfC90NORLtOisICN&q=85&s=3524a3290d04babf987ca30dfb17a09a" alt="MCP data flow" width="2160" height="600" data-path="images/mcp-dataflow.png" />

<Note>
  This is an **advanced feature** for people comfortable with the command line / AI coding tools. If you only make resumes in the browser, you don't need this page.
</Note>

It runs a small program over **MCP** (a standard protocol), spawned locally by your AI tool. Two steps: configure credentials once, then register it with each tool.

<Steps>
  <Step title="Create a Personal Access Token (PAT)">
    In the web app, **Settings → Personal Access Tokens → Create**, and copy the `mr_pat_…` value (shown once). See [Settings](/en/guide/settings).
  </Step>

  <Step title="Configure the MCP server">
    ```bash theme={null}
    npx -y @magic-resume/mcp config set \
      --api-url "https://api.magic-resume.cn/api" \
      --pat "mr_pat_xxx"
    ```

    This writes `~/.magic-resume/mcp.json`. If you run the backend locally, point `--api-url` at `http://localhost:3110/api` (the Core gateway).
  </Step>

  <Step title="Register with your AI tool">
    <Tabs>
      <Tab title="Claude Code">
        ```bash theme={null}
        claude mcp add magic-resume -- npx -y @magic-resume/mcp mcp
        ```

        This adds Magic Resume to the user-scoped MCP list. Restart Claude Code so it picks up the new server.
      </Tab>

      <Tab title="Cursor">
        Open `~/.cursor/mcp.json` (create if missing) and add:

        ```json theme={null}
        {
          "mcpServers": {
            "magic-resume": {
              "command": "npx",
              "args": ["-y", "@magic-resume/mcp", "mcp"]
            }
          }
        }
        ```
      </Tab>

      <Tab title="Manual / Windsurf">
        Any MCP client that supports stdio servers can spawn it:

        ```bash theme={null}
        npx -y @magic-resume/mcp mcp
        ```

        The server speaks the standard MCP protocol over stdin/stdout — no flags needed.
      </Tab>
    </Tabs>
  </Step>

  <Step title="Try it">
    In your AI tool, ask: "List my Magic Resume resumes." The model should call the tool and return your list. If it errors, run `npx -y @magic-resume/mcp config print` and check the API URL and PAT.
  </Step>
</Steps>

<Warning>
  AI tools edit resumes the same safe way — **schema-aware, via JSON Patch** — never full rewrites, never invented fields. But the PAT lets it read/write all your resumes, so keep it safe.
</Warning>

<Note>
  Curious how MCP works internally and which tools exist? That's developer material — see the [MCP section of the Developer Handbook](/en/mcp).
</Note>