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

# Use with AI (MCP)

> Drive the ScripX API docs from AI tools: plain-text endpoints for LLMs, per-page copy and open-in-assistant actions, and the machine-readable OpenAPI contract your agent can code against.

These docs are built to be driven by AI. Point your assistant at them and it can answer integration questions with the exact contract, not a guess.

## Ask AI on any page

Every page has a contextual menu next to its title: copy the page as Markdown, view it as plain text, open it directly in ChatGPT or Claude with the page pre-loaded as context, or copy the MCP server URL and one-click connect to Cursor or VS Code.

## Connect via MCP

These docs run a hosted [Model Context Protocol](https://modelcontextprotocol.io) server, generated from the same pages and OpenAPI contract you see here. Point any MCP-capable client at it and your agent can search the docs and read the API surface as live tools, always current with what ships.

| Endpoint   | URL                                         |
| ---------- | ------------------------------------------- |
| MCP server | `https://docs.scripxhq.com/mcp`             |
| Discovery  | `https://docs.scripxhq.com/.well-known/mcp` |

Add it as a remote server in your client's MCP config:

```json theme={"dark"}
{
  "mcpServers": {
    "scripx-docs": {
      "url": "https://docs.scripxhq.com/mcp"
    }
  }
}
```

For Cursor and VS Code, use the **Connect** action in any page's contextual menu to install it in one click, or **Copy MCP server URL** to paste it wherever your tool expects a remote MCP endpoint. The server is read-only: it exposes the public docs and the `/v1` contract, never your keys or data.

## Plain text for LLMs

When deployed, the docs serve the standard LLM endpoints:

* `/llms.txt`, an index of every page with one-line summaries.
* `/llms-full.txt`, the entire docs corpus as one plain-text file.

## The machine-readable contract

The single best artifact to hand an agent is the OpenAPI document, served publicly with no key:

```bash theme={"dark"}
curl https://api.scripxhq.com/v1/openapi.json
```

It is hand-maintained against what the service actually enforces, the same auth header, the same scopes, the same endpoints, and it is the source these reference pages render from.

<Tabs>
  <Tab title="Claude Code">
    ```bash theme={"dark"}
    # Give Claude Code the contract and let it write the client.
    curl -s https://api.scripxhq.com/v1/openapi.json -o scripx-openapi.json
    claude "Generate a typed client for scripx-openapi.json with retry on 429 honouring Retry-After"
    ```
  </Tab>

  <Tab title="Cursor">
    ```json theme={"dark"}
    // Add the docs to Cursor's context (Settings → Indexing & Docs).
    { "docs": ["https://api.scripxhq.com/v1/openapi.json"] }
    ```
  </Tab>

  <Tab title="ChatGPT">
    Paste the OpenAPI JSON into the conversation, or use the page-level "Open in ChatGPT" action, and ask for an integration plan: quote, order with idempotency, webhook verification.
  </Tab>
</Tabs>

<Note>
  The docs and the OpenAPI document are **read-only** context. Your API key is yours: it is never part of the docs, the contract, or any example on this site. Keys go in your server's environment, not in a prompt.
</Note>

## Related

* [Quick start](/quick-start), the five calls an agent should wire first.
* [Webhooks](/webhooks), signature verification your agent must not skip.