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

# Build with AI agents

> Give coding agents and AI assistants everything they need to use the API correctly.

These docs are built to be read by agents as well as people.

<CardGroup cols={2}>
  <Card title="llms.txt" icon="file-lines" href="https://docs.quoindata.com/llms.txt">
    An index of every page, with a one-line summary each.
  </Card>

  <Card title="llms-full.txt" icon="book" href="https://docs.quoindata.com/llms-full.txt">
    The complete docs in one plain-text file.
  </Card>

  <Card title="OpenAPI spec" icon="brackets-curly" href="https://portal.quoindata.com/api/openapi.json">
    Machine-readable endpoints, parameters and response schemas. `x-billable` marks billable operations.
  </Card>

  <Card title="Docs MCP server" icon="plug" href="#connect-the-docs-mcp-server">
    Let your assistant search these docs directly.
  </Card>
</CardGroup>

Every page is also available as Markdown: add `.md` to its URL. The menu at the top of each page copies it, or opens it in ChatGPT, Claude, Cursor and others.

## Connect the docs MCP server

The server at `https://docs.quoindata.com/mcp` exposes a search tool over these docs.

<CodeGroup>
  ```bash Claude Code theme={null}
  claude mcp add --transport http quoin-docs https://docs.quoindata.com/mcp
  ```

  ```json Cursor (.cursor/mcp.json) theme={null}
  {
    "mcpServers": {
      "quoin-docs": { "url": "https://docs.quoindata.com/mcp" }
    }
  }
  ```
</CodeGroup>

## Prompt for coding agents

Paste this into your agent to set up the API in a project:

```text theme={null}
Use the Quoin Property API, current version only (docs: https://docs.quoindata.com/llms-full.txt,
OpenAPI: https://portal.quoindata.com/api/openapi.json). Ignore anything about v2 or v3.
Read the API key from the QUOIN_KEY environment variable; never hard-code it.
Follow the rules under "Rules for agents" in the docs.
```

## Rules for agents

1. **Use only the current version, v4.** Ignore anything that mentions `/api/v2` or `/api/v3`; those versions are retired. See [Versioning](/api-basics/versioning).
2. **Resolve before you read.** Use free `find` and `connected` to get the right q\_id before any billable call. If `find` returns several matches, ask the user to choose.
3. **Prefer `summary`, then `records` with `kind`.** Don't page through every record when one summary or one kind answers the question.
4. **Never loop billable calls over large result sets** without the user's approval. Every page of `records` or `search` is billed.
5. **Send an `Idempotency-Key`** on billable calls, and reuse it only to retry the same call.
6. **Discover fields with `catalog`** before writing a search filter. Field names are case-sensitive and differ by source.
7. **Handle errors by `error.code`.** Retry `503` with backoff and the same key; honour `Retry-After` on `429`; fix the request on `400`.
8. **Keep citations.** Show `citation.source` and `citation.as_of` with any value you present.
9. **Treat owner matches as text, not identity.** Confirm with the address. See [Owner lookup](/guides/owner-lookup).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.