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

# Docs as MCP

> Point your LLM at these docs and vibe-code automations against your Magic Meal Kits server.

This documentation site is itself available as an **MCP server** and as `llms.txt`, so
an LLM can read and search the entire API reference while it writes your automation
code.

| Resource     | URL                                            |
| ------------ | ---------------------------------------------- |
| MCP endpoint | `https://docs.magicmealkits.com/mcp`           |
| LLM index    | `https://docs.magicmealkits.com/llms.txt`      |
| Full text    | `https://docs.magicmealkits.com/llms-full.txt` |

<Note>
  The docs MCP serves **documentation content** — it lets the model read and search this
  reference. It does **not** proxy your server. The code the model writes calls your own
  deployment's URL with your own API key.
</Note>

## Add it to your client

**Claude Code**

```bash theme={null}
claude mcp add --transport http mmk-docs https://docs.magicmealkits.com/mcp
```

**Claude Desktop / Cursor** — add an MCP server entry pointing at
`https://docs.magicmealkits.com/mcp`.

## Then vibe-code

Ask your model something like:

> Using the Magic Meal Kits docs, write a script that pulls a YouTube transcript and
> posts a summary to my Notion database. My server is `https://magic-meal-kits-xxxxx.run.app`.

The model reads this reference over MCP, finds the right endpoints, and writes code
that calls *your* deployment.

<Info>
  **Skeleton page.** Add concrete recipes (Notion sync, Threads analytics, transcription
  pipeline) once the API Reference pages exist.
</Info>