How the TestDriver agent behaves on GitHub issues, pull requests, and @mentions
中文简介正在准备中。当前显示原文说明。
Add TestDriver to an existing project by hand.
中文简介正在准备中。当前显示原文说明。
查看安装方法安装前可以查看交给 AI 代理的指令内容。
Add TestDriver to an existing project without the init scaffold. This is useful when you already have a Vitest setup or want full control over each file.
<Card
title="Get an API Key"
icon="user-plus"
href="https://console.testdriver.ai/settings"
arrow
horizontal
>
Start with 60 free device minutes, no credit card required.
</Card>
Save the key in a `.env` file at your project root. The SDK loads it automatically:
```bash .env
TD_API_KEY=your_api_key
```
Add `.env` to `.gitignore` so the key is not committed.
</Step>
<Step title="Install dependencies">
Install Vitest and TestDriver as dev dependencies:
```bash
npm install --save-dev vitest testdriverai
```
<Note>
TestDriver requires Node.js 20.19 or later and only runs on [Vitest](https://vitest.dev). Jest, Mocha, and other runners are not supported.
</Note>
</Step>
<Step title="Configure Vitest">
Create `vitest.config.js` at your project root (or add these settings to an existing one):
```js vitest.config.js
import { defineConfig } from 'vitest/config';
import TestDriver from 'testdriverai/vitest';
export default defineConfig({
test: {
// Sandboxes take time to boot and tear down. Both values are required.
testTimeout: 300000,
hookTimeout: 300000,
reporters: ['default', TestDriver()],
setupFiles: ['testdriverai/vitest/setup'],
},
});
```
- `TestDriver()` reporter uploads results, recordings, and screenshots to the console.
- `setupFiles` registers the hooks that connect each test to a sandbox and clean it up.
- Without `hookTimeout`, cleanup fails at Vitest's default 10-second limit.
</Step>
<Step title="Write a test">
Create `tests/search.test.js`:
```js tests/search.test.js
import { test, expect } from 'vitest';
import { TestDriver } from 'testdriverai/vitest/hooks';
test('search shows results', async (context) => {
// Connects to a sandbox and records the session
const testdriver = TestDriver(context);
// Launch Chrome at a URL
await testdriver.provision.chrome({ url: 'https://duckduckgo.com' });
// Locate elements by describing them
const searchBox = await testdriver.find('search input field');
await searchBox.click();
// Type into the focused element and submit
await testdriver.type('testdriver.ai');
await testdriver.pressKeys(['enter']);
// Ask a yes/no question about the screen
const result = await testdriver.assert('search results are displayed');
expect(result).toBeTruthy();
});
```
If your project does not have `"type": "module"` in `package.json`, name the file `search.test.mjs` instead.
</Step>
<Step title="Run the test">
```bash
npx vitest run
```
A sandbox starts, Chrome opens, and a live preview appears in your browser. When the run finishes, open the `TESTDRIVER_RUN_URL` printed at the end of the output to see the recording and step-by-step screenshots.
</Step>
</Steps>
If you want to write tests with an AI assistant, connect the TestDriver agent and MCP server. You can do this without re-scaffolding the project:
npx testdriverai init --client cursor,claude-code,vscode --no-sample-test
See Configure Your Agent for the manual configuration of each client.
还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问,或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
How the TestDriver agent behaves on GitHub issues, pull requests, and @mentions
中文简介正在准备中。当前显示原文说明。
Execute natural language tasks using AI
中文简介正在准备中。当前显示原文说明。
Make AI-powered assertions about screen state
中文简介正在准备中。当前显示原文说明。
Deploy TestDriver on your AWS infrastructure using CloudFormation
中文简介正在准备中。当前显示原文说明。
Speed up tests with screenshot-based caching
中文简介正在准备中。当前显示原文说明。
How TestDriver learns your app and caches what it discovers for instant, deterministic replays
中文简介正在准备中。当前显示原文说明。