# Build with an AI agent

> Point Cursor, Claude Code, Copilot or ChatGPT at these docs.

Source: https://docs.onari.xyz/docs/build-with-ai

These docs are written so coding agents can follow them.

- **`https://docs.onari.xyz/llms.txt`**: an index of every page, for agents that support it.
- **`https://docs.onari.xyz/llms-full.txt`**: the whole site as one Markdown file.
- **Any page as Markdown:** add `.md` to its URL, for example `/docs/webhooks.md`. Every page also has **Copy as Markdown**.

## A prompt that works

```text
Integrate Onari (https://docs.onari.xyz/llms-full.txt) into this codebase.

1. Find where these actions are confirmed on the server: <list your actions, e.g. order paid, transfer settled>.
   After each one, POST it to https://api.onari.xyz/v1/events with the user's internal ID as player_id,
   the amount in minor units, and our transaction ID as idempotency_key. Don't block the user's request on it;
   log and retry on failure. Never send names, phone numbers or emails.
2. Add POST /onari/webhooks that verifies Standard Webhooks signatures with ONARI_WEBHOOK_SECRET using the
   raw body, and for reward.issued (not vouchers) pays the player exactly once using idempotency_key with a
   unique constraint. Reply 200 fast.
3. Add a server endpoint that calls POST /v1/hub/sessions for the signed-in user and returns the url,
   and open it in a WebView from our Rewards screen.
Use ONARI_SECRET_KEY from the environment. Write tests for the webhook handler.
```