e2e-testing（端到端测试）

AI E2E Testing — 8 Platforms, Zero Test Code

Give your AI agent eyes and hands inside any running app.

flutter-skill is an MCP server that connects AI agents to running apps. The agent can see screenshots, tap elements, enter text, scroll, navigate, inspect UI trees, and verify state — all through natural language.

Supported Platforms

Test scorecard: 562/567 (99.1%) across all 8 platforms.

Install

# npm (recommended)
npm install -g flutter-skill

# Homebrew
brew install ai-dashboad/flutter-skill/flutter-skill

# Or download binary from GitHub Releases

MCP Configuration

Add to your AI agent's MCP config (Claude Desktop, Cursor, Windsurf, OpenClaw, etc.):

{
  "mcpServers": {
    "flutter-skill": {
      "command": "flutter-skill",
      "args": ["server"]
    }
  }
}

OpenClaw

If using OpenClaw, add to your gateway config under mcp.servers:

mcp:
  servers:
    flutter-skill:
      command: flutter-skill
      args: ["server"]

Quick Start

1. Initialize your app (one-time)

cd /path/to/your/app
flutter-skill init

Auto-detects project type and patches your app with the testing bridge.

2. Launch and connect

flutter-skill launch .

3. Test with natural language

Tell the agent what to test:

"Test the login flow — enter admin@test.com and password123, tap Login, verify Dashboard appears"

The agent will automatically:

1. screenshot() → see the current screen
2. inspect_interactive() → discover all tappable/typeable elements with semantic refs
3. tap(ref: "button:Login") → tap using stable semantic reference
4. enter_text(ref: "input:Email", text: "admin@test.com") → type into field
5. wait_for_element(key: "Dashboard") → verify navigation
6. screenshot() → confirm final state

Available MCP Tools

Core Actions

Inspection (v0.8.0+)

Text Manipulation

App Control

Semantic Refs (v0.8.0)

inspect_interactive returns elements with stable semantic reference IDs:

button:Login          → Login button
input:Email           → Email text field
toggle:Dark Mode      → Dark mode switch
button:Submit[1]      → Second Submit button (disambiguated)

Format: {role}:{content}[{index}]

7 roles: button, input, toggle, slider, select, link, item

Use refs for reliable element targeting that survives UI changes:

tap(ref: "button:Login")
enter_text(ref: "input:Email", text: "test@example.com")

Testing Workflow

Basic Flow

screenshot() → inspect_interactive() → tap/enter_text → screenshot() → verify

Comprehensive Testing

"Explore every screen of this app. Test all buttons, forms, navigation, and edge cases. Report any bugs you find."

The agent will systematically:

• Navigate every screen via tab bars, menus, links
• Interact with every interactive element
• Test form validation (empty, invalid, valid inputs)
• Test edge cases (long text, special characters, emoji)
• Verify navigation flows (forward, back, deep links)
• Take screenshots at each step for verification

Example Prompts

Quick smoke test:

"Tap every tab and screenshot each page"

Form testing:

"Fill the registration form with edge case data — emoji name, very long email, short password — and verify error messages"

Navigation:

"Test the complete user journey: sign up → create post → like → comment → delete → sign out"

Accessibility:

"Check every screen for missing labels, small tap targets, and contrast issues"

Tips

1. Always start with screenshot() — see before you act
2. Use inspect_interactive() to discover elements — don't guess at selectors
3. Prefer ref: selectors — more stable than text or coordinates
4. wait_for_element() after navigation — apps need time to transition
5. Screenshot after every action — verify the expected effect
6. Use press_key for keyboard shortcuts — test keyboard navigation

Links

• GitHub
• npm
• Documentation
• Demo Video
• pub.dev
• VSCode Extension