# AI agent setup

> One prompt that sets up Claude Code, Codex, Cursor and other agents to work with the BILLZ API.

Need answers about your company’s data rather than code? See [Connect to Claude and ChatGPT](/en/mcp/).

Copy the phrase below and send it to your agent: Claude Code, Codex, Cursor, Windsurf, GitHub Copilot or any other agent that can download pages by link. The same phrase is copied by the “Prompt for an AI agent” button on the home page.

```text
Read and follow the instructions for setting up an AI agent to work with the BILLZ public API: https://docs.billz.io/en/agent-setup/prompt.md
```

## What the agent will do

The instructions are published in plain form: [/en/agent-setup/prompt.md](/en/agent-setup/prompt.md). The agent:

1. Downloads the [OpenAPI specification](/en/openapi.yaml) into the project (`docs/billz-openapi.yaml`) and learns where to find the documentation: [/llms.txt](/llms.txt) and the markdown versions of the pages.
2. Adds a “BILLZ API” section to the project rules file (`AGENTS.md` or `CLAUDE.md`): authentication and token refresh, the `Billz-Response-Channel: HTTP` header for asynchronous methods, behavior on `408` and `429`, error formats.
3. Asks whether it may call the production API: every write operation is executed in your real company.
4. Reports what it has set up and suggests the [first request](/en/start/first-request/) and [choosing a scenario](/en/start/scenarios/).

The prompt installs nothing and changes nothing outside your project: only a copy of the specification and a section in the rules file.

## If the agent cannot follow links

Open [/en/agent-setup/prompt.md](/en/agent-setup/prompt.md), copy the whole text and paste it to the agent together with the specification.