AI coding assistants

This documentation publishes files that AI coding assistants (Claude Code, Cursor, GitHub Copilot, ChatGPT and others) can read directly, so they use the current endpoints, field names and signing rules instead of guessing.

File	Content	Use it to
/llms.txt	Index of every page with links (llms.txt format)	Let the assistant pick only the relevant pages
/llms-full.txt	Every page, Thai and English, as one Markdown file	Give the assistant the full context at once
/openapi.json	OpenAPI 3.0 description of the Partner API	Generate clients, types or mocks, and validate requests and responses

In the console, the AI integration guide menu opens llms-full.txt for you.

Never paste your secret into an AI tool

Do not put your API secret in a chat, a prompt, a file you share with an assistant, or code it generates. Code must read it from an environment variable (such as PARTNER_API_SECRET). If you pasted it by mistake, create a new secret in the console right away.

How to use them

1. Point the assistant at https://docs.example.com/llms-full.txt (paste the URL if the tool can browse, or download the file into your repository and reference it).
2. For generated types or clients, use https://docs.example.com/openapi.json.
3. State your language, framework and what you need.
4. Review the result, especially the signing function: check it against the test vectors before calling the API.

Example prompt:

Read https://docs.example.com/llms-full.txt (Partner API section).
Write a TypeScript module for Node.js 20 that:
- signs requests exactly as described (X-Api-Key, X-Timestamp, X-Nonce, X-Signature),
  reading PARTNER_API_KEY and PARTNER_API_SECRET from environment variables,
- creates a pay-in with the payer's bank account inline (payer: { bankCode, accountNumber, accountName }),
  and returns checkoutUrl, amount, payerId and payTo (incl. payTo.qrPayload),
- exposes an Express route that verifies webhook signatures on the raw body
  and handles payin.succeeded idempotently (dedupe on data.id + data.status).
Add a unit test that checks the signer against the documented test vectors.

Things to watch

• These files are regenerated whenever the documentation changes. Fetch them again after updates instead of relying on an old copy.
• If the assistant uses a field, header or endpoint that is not in the documentation, trust the documentation and openapi.json.
• Do not let AI tools call the API automatically with production keys: pay-ins and payouts move real money.