Connect Claude Code, Cursor, and other AI tools to your API specs

Set it up in 4 steps

1. 1Sign in with Google.
2. 2Open Connect AI and copy your MCP URL.
3. 3Add the URL to your AI tool. For Claude Code, run the command below.
4. 4Ask your AI about your APIs in plain language.

Claude Code

claude mcp add --transport http apidrawer "<your-mcp-url>"

Other MCP clients

Tools that accept a remote MCP server URL usually take a JSON config like this. Check your tool's docs for where the file goes.

{
  "mcpServers": {
    "apidrawer": {
      "url": "<your-mcp-url>"
    }
  }
}

What your AI can do

• list_apis: list the APIs in your drawer, filtered by tag, category, or status.
• get_api: read an API's endpoint, auth type, and field descriptions.
• list_samples: read sample responses, primary sample first.
• create_api: add a new API with its fields.
• update_api: fill in or fix fields and descriptions.
• list_combos and create_combo: see and save groups of APIs.
• add_sample: save a real response as a sample.

Example prompts

• What weather APIs do I have in my drawer? Show me the response fields.
• Add the MTA bus arrivals API from this doc, with every request parameter and response field described.
• Save this response as a sample for BUS_ARRIVALS_GET. Mask any personal data first.
• Using the APIs in my drawer, build a Next.js page that shows the forecast and the next bus.

Keep your URL private