English

English
简体中文

See interfaces like a humanE2E testing in natural language

Midscene.js helps developers bring AI to E2E testing with a vision-based GUI Agent and an extensible testing framework for web, mobile, and desktop apps.

Documentation Showcases

15k+Github Stars

#2No.2 in Github trending

@midscene/test + YAML test cases

beforeEach:
  - gotoUrl: https://your-shop.example

cases:
  - name: "Find headphones under $100"
    steps:
      - aiAct: "Search for headphones and filter the results to prices under $100"
      - aiWaitFor: "The filtered search results are visible"
      - aiAssert: "There are products in the search results, and each costs less than $100"

Vitest + JavaScript

import { test } from 'vitest';
import { PlaywrightAgent } from '@midscene/web/playwright';

// Model configuration, page setup, and cleanup omitted
test("Find headphones under $100", async () => {
  const agent = new PlaywrightAgent(page);
  await agent.aiAct("Search for headphones and filter the results to prices under $100");
  await agent.aiWaitFor("The filtered search results are visible");
  await agent.aiAssert("There are products in the search results, and each costs less than $100");
}, 180_000);

Locate and act from screenshots

Locate elements and operate interfaces from screenshots, without maintaining selectors or annotations. Works with custom interfaces such as Canvas.

Multiple model options

Choose from Doubao, DeepSeek, Qwen, GPT, Gemini, Kimi, and more to fit your tasks and budget.

Low model API cost

In this AppControlBench run, Doubao Seed 2.1 Turbo passed 58 of 60 tasks at a total model API cost of $0.59.

View benchmark report

Verify what users see

Describe expected colors, layouts, and states in natural language. Visual assertions check what the interface actually shows.

Web
@midscene/web
Playwright · Puppeteer · Desktop Chrome
Playwright Puppeteer Bridge Mode
Android
@midscene/android
Devices / emulators · ADB
Get started
iOS
@midscene/ios
Devices / simulators · WebDriverAgent
Get started
HarmonyOS
@midscene/harmony
HarmonyOS NEXT · HDC
Get started
Desktop
@midscene/computer
macOS · Windows · Linux, with remote RDP
Get started
Custom interfaces
@midscene/core
Connect through screenshots and actions
Extension guide

@midscene/test

YAML for test cases TypeScript for extensions

Organize cases in natural language

Combine UI actions, assertions, and business Nodes in YAML to describe your test flows.

Make business operations reusable

Implement data setup, API calls, and cleanup in TypeScript Nodes that different cases can reuse.

Generate a Node Spec automatically

Export registered Nodes and input schemas as Markdown so developers and AI agents can discover project capabilities and maintain tests together.

Explore Midscene Test

YAML cases

TypeScript extensions

Node Spec

# cases/users.yaml
# Example: prepare user data, then search and verify through the UI. Configure a model and Web environment first, and replace the example URL.

beforeEach:
  - user.create: { name: Alice }
  - gotoUrl: https://your-app.example/users

cases:
  - name: Find a new user
    steps:
      - aiAct: Search for the user Alice
      - aiAssert: The search results show the user Alice

// nodes.ts
// user.create matches the YAML step. Implement userService with your business APIs to create and delete test users.

import { defineNode, z } from '@midscene/test';
import { userService } from './user-service';

export const createUser = defineNode({
  name: 'user.create',
  description: "Create a test user and clean up after the case",
  inputSchema: z.strictObject({ name: z.string() }),
  async execute({ input, onTeardown }) {
    const user = await userService.create(input);
    onTeardown(() => userService.remove(user.id));
  },
});

<!-- midscene-node-spec.md
Generated by pnpm exec midscene-test nodes. This excerpt describes user.create and matches its TypeScript definition. -->

### `user.create`

Create a test user and clean up after the case

**String shorthand:** Not supported by this Node.

#### Input Schema

```json
{
  "type": "object",
  "properties": { "name": { "type": "string" } },
  "required": ["name"]
}
```

<!-- Remaining schema fields and Node descriptions omitted. -->

Rich, flexible APIs

Bring autonomous flows, precise actions, visual assertions, and data extraction into your existing test stack.

Replay runs, debug interactively

Review screenshots, AI decisions, and execution results in reports. Try and refine instructions in the Playground.

Let AI coding agents operate your UI

Give AI coding agents Midscene Skills to operate apps, verify interfaces, and inspect execution results.

93.1%Pass@1 93.1% · Pass@3 97.4%

Write your first E2E test in natural language

Get started Explore examples

See interfaces like a humanE2E testing in natural language

Describe steps and expectations in natural language

@midscene/test + YAML test cases

Vitest + JavaScript

Vision-driven Low-cost testing

Locate and act from screenshots

Multiple model options

Low model API cost

Verify what users see

One API, across platforms

Web

Android

iOS

HarmonyOS

Desktop

Custom interfaces

YAML for test cases TypeScript for extensions

Organize cases in natural language

Make business operations reusable

Generate a Node Spec automatically

A complete toolkit for AI-native testing

Rich, flexible APIs

Replay runs, debug interactively

Let AI coding agents operate your UI

Benchmarks

AndroidWorld Benchmark

MobileWorld Benchmark

AppControlBench Benchmark

Who is using Midscene

Write your first E2E test in natural language