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

# Use the docs from your AI assistant

> Pull these docs into Claude Code, Cursor, ChatGPT, or any MCP-capable agent — three integration paths, no setup on our side.

These docs are built to be consumed by AI coding agents directly, so you can ask questions like *"how do I send a sticker with zapo?"* or *"what events fire during pairing?"* and get answers grounded in the current reference — not stale model memory.

<Note>
  This page is about the **documentation** MCP — exposing the zapo.to pages to your assistant. It is **not** the [`@zapo-js/mcp-server`](/en/dev-tools#mcp-server) dev tool, which drives a live `WaClient` from an agent. Different goals, different setup, and you can use them together.
</Note>

## Pick a path

<CardGroup cols={3}>
  <Card title="One-click contextual" icon="hand-pointer">
    Use the **Copy / Open in…** menu at the top of any page to send its content to Claude, ChatGPT, or Cursor. No setup.
  </Card>

  <Card title="MCP server" icon="plug">
    Add `https://zapo.to/mcp` as an MCP server in your client — the agent then searches and fetches docs pages on demand.
  </Card>

  <Card title="llms.txt" icon="file-lines">
    Plain-text bundles at [`/llms.txt`](https://zapo.to/llms.txt) (index) and [`/llms-full.txt`](https://zapo.to/llms-full.txt) (full corpus) for tools that don't speak MCP.
  </Card>
</CardGroup>

***

## Contextual buttons

Every page on [zapo.to](https://zapo.to) has a menu near the title that lets you:

* **Copy page as Markdown** — paste straight into any chat.
* **View as Markdown** — see the raw source.
* **Open in ChatGPT / Claude / Cursor** — opens the assistant with the current page pre-loaded as context.
* **Connect MCP** — copy the MCP URL for any other client.

Best for a one-off question about a single page.

***

## MCP server

For sustained work — building an integration, debugging an event, planning a migration from Baileys — register the docs as a long-lived MCP server. The agent then searches and fetches pages itself as needed, without you copy-pasting.

The endpoint is:

```text theme={null}
https://zapo.to/mcp
```

### Claude Code

```bash theme={null}
claude mcp add zapo-docs --scope user --transport http https://zapo.to/mcp
```

### Cursor

Add to `~/.cursor/mcp.json` (or the workspace-local `.cursor/mcp.json`):

```json theme={null}
{
  "mcpServers": {
    "zapo-docs": {
      "url": "https://zapo.to/mcp"
    }
  }
}
```

### Other clients

Any MCP-compatible client (Windsurf, Zed, Continue, custom agents built on the [MCP spec](https://modelcontextprotocol.io)) can connect to the same URL — refer to your client's docs for the exact registration syntax. The server exposes search and page-fetch tools.

<Tip>
  Pair the docs MCP with the [`@zapo-js/mcp-server`](/en/dev-tools#mcp-server) dev tool to get an agent that can both **read** the docs and **drive** a live `WaClient` — handy for *"explain what this event means, then trigger it on a real session"* workflows.
</Tip>

***

## llms.txt

For tools that don't support MCP yet, the same corpus is published as static text following the [llms.txt](https://llmstxt.org) convention:

| File                                              | Contents                                                                                        |
| ------------------------------------------------- | ----------------------------------------------------------------------------------------------- |
| [`/llms.txt`](https://zapo.to/llms.txt)           | Index of every page with titles and one-line summaries — small enough to paste into any prompt. |
| [`/llms-full.txt`](https://zapo.to/llms-full.txt) | Full page bodies, concatenated. Large; best for one-shot context windows or RAG ingestion.      |

Both files are regenerated on every deploy, so they always reflect the live docs.