# Quickstart
> Send one message and sign in once, and Beatra is available in your agent. The flow below is the same in any agent that can install a Skill and run Python 3.10 or newer.

Steps last verified: 2026-11-09

You will install the Beatra Skill in your agent, sign in once in the browser, and get a first result. The flow is the same in Claude Code, Cursor, Codex, WorkBuddy, and any other agent that can install a Skill and run Python 3.10 or newer. Installing and signing in are free; creating media uses Beatra credits. To compare the Skill with MCP and the REST API first, see [Choose how to use Beatra](https://docs.beatra.cn/en/docs/choose-integration).

### Send the install message to your agent

```text title="Send to your agent"
Install Beatra and tell me when I need to sign in: https://beatra.ai/install.md
```

Your agent identifies the platform, finishes the setup it can do on its own, and tells you if a platform setting needs your help. What it installs, and how to update or reconnect later, is on [Install the Beatra Skill](https://docs.beatra.cn/en/docs/install-skill).

**Checkpoint**

- [x] The agent confirms the install is complete.

### Sign in once in the browser

When Beatra opens in the browser, sign in and approve access once. That one sign-in covers image, video, music, and speech; switching media types does not need another authorization. After that, start a new agent session so Beatra loads.

**Checkpoint**

- [x] The browser page shows "Approved", then "Connected" once your agent picks up the sign-in.
- [x] The agent reports that Beatra is ready.

### Start creating

Describe the result in ordinary language:

```text title="Send to your agent"
Create a warm editorial illustration for this landing page.
```

Longer work keeps moving in the background, and the agent returns the result when it is ready. Creating media uses credits.

**Checkpoint**

- [x] The agent confirms a Beatra task was created.
- [x] When it finishes you receive the artifact and the charge. A task still waiting in line, or still generating, is normal. The agent keeps checking progress.

### View it in the console or on your phone

The work appears on the [console assets page](https://console.beatra.cn/assets). You can also scan with WeChat and open it in the [Beatra mini program](https://www.beatra.cn/en/mini-program?utm_source=docs&utm_medium=quickstart).

**Checkpoint**

- [x] The console's Work page lists the result.

## If a step does not work [#if-a-step-fails]

Send the same install message again. The agent keeps the steps that are already correct and continues from the part that still needs attention. For a specific message:

- The approval page, the sign-in code, or a sign-in that has ended: [Troubleshooting: sign-in fails](https://docs.beatra.cn/en/docs/troubleshooting#sign-in), and [Sign-in troubleshooting](https://docs.beatra.cn/en/docs/sign-in-help) for the WeChat QR code and verification codes.
- Insufficient credits: [Troubleshooting: not enough credits](https://docs.beatra.cn/en/docs/troubleshooting#credits).
- A task that keeps running or ends in an error: [Troubleshooting: slow tasks and timeouts](https://docs.beatra.cn/en/docs/troubleshooting#timeouts).

## Building your own service

To integrate Beatra in your own application, use REST/OpenAPI:

- [REST · API reference](https://docs.beatra.cn/en/docs/api-reference) — Every v1 endpoint and the machine-readable contract.
- [Before you start · Authentication & request conventions](https://docs.beatra.cn/en/docs/authentication) — API keys, shared headers, and idempotency.
- [Mechanics · Async tasks](https://docs.beatra.cn/en/docs/mechanisms/async-tasks) — The lifecycle from submit to a terminal state.