AI 에이전트가 자연어를 통해 실행 중인 모바일, 웹, 데스크톱 앱을 직접 조작하고 E2E 테스트를 수행할 수 있게 합니다.
다른 이름 ai-dashboad/flutter-skill
- 스타
- 407
- Tools
- —
- 다운로드, 30일
- —
- 최신 버전
- 0.9.36
- 마지막 커밋
- 2026년 9월 1일
- 라이선스
- MIT
flutter-skill MCP란?
flutter-skill은(는) MCP server입니다. AI 에이전트가 자연어를 통해 실행 중인 모바일, 웹, 데스크톱 앱을 직접 조작하고 E2E 테스트를 수행할 수 있게 합니다.
프로젝트 README에서 생성된 요약
flutter-skill MCP server 정보
모바일, 웹, 데스크톱 애플리케이션을 개발하는 개발자는 flutter-skill을 사용하여 AI 에이전트가 자연어로 실행 중인 소프트웨어와 상호작용할 수 있는 기능을 부여할 수 있습니다. 이 MCP server를 Claude Desktop, Cursor 또는 Windsurf와 같은 호환 클라이언트에 연결하면, 에이전트가 화면을 보고 버튼 탭, 텍text 입력, 스크롤 등의 동작을 수행할 수 있습니다. 이 설정을 통해 수동 테스트 코드를 작성하거나 Page Object 또는 취약한 XPath를 만들 필요가 없습니다.
서버는 두 가지 주요 모드로 작동합니다. IDE에 직접 통합되는 MCP stdio 모드와 CLI 자동화 및 CI/CD 파이프라인을 위한 HTTP serve 모드가 있습니다. Flutter, React Native, Electron, Android, iOS를 포함한 10가지 플랫폼을 지원합니다. 비용이 많이 드는 스크린샷에 의존하는 대신 accessibility tree를 사용하여 인터페이스의 의미론적 요약을 제공하므로 토큰 사용량을 크게 줄일 수 있습니다. 이를 통해 복잡한 애플리케이션 흐름을 더 빠르고 효율적으로 자율 탐색 및 테스트할 수 있습니다.
활용 사례
- 1자율 E2E 테스트스크립트를 작성하지 않고 AI 에이전트에게 특정 사용자 흐름을 탐색하고 결과를 검증하도록 요청하세요.
- 2자동 버그 발견AI가 애플리케이션 전체를 탐색하여 기능적 오류를 찾아내고 보고하도록 지시하세요.
- 3안정성을 위한 퍼즈 테스트애플리케이션에 무작위 동작을 실행하여 잠재적인 충돌이나 예외 케이스를 감지하세요.
- 4효율적인 UI 상호작용accessibility tree를 사용하여 탭이나 타이핑과 같은 고속 상호작용을 수행하세요.
프로젝트의 README를 바탕으로 작성되었습니다. 모든 활용 사례는 해당 내용의 인용문과 대조하여 확인합니다.
flutter-skill MCP 설치 방법
형식 확인: 2026년 11월 10일Cursor에 추가Goose에 추가Claude Code에 flutter-skill MCP 설치
$ claude mcp add flutter-skill -- npx -y flutter-skillCursor에 flutter-skill MCP 설치
~/.cursor/mcp.json
{
"mcpServers": {
"flutter-skill": {
"command": "npx",
"args": [
"-y",
"flutter-skill"
]
}
}
}VS Code에 flutter-skill MCP 설치
.vscode/mcp.json
{
"servers": {
"flutter-skill": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"flutter-skill"
]
}
}
}VS Code CLI에 flutter-skill MCP 설치
$ code --add-mcp '{"name":"flutter-skill","type":"stdio","command":"npx","args":["-y","flutter-skill"]}'Claude Desktop에 flutter-skill MCP 설치
claude_desktop_config.json
{
"mcpServers": {
"flutter-skill": {
"command": "npx",
"args": [
"-y",
"flutter-skill"
]
}
}
}Codex에 flutter-skill MCP 설치
$ codex mcp add flutter-skill -- npx -y flutter-skillCodex config.toml에 flutter-skill MCP 설치
~/.codex/config.toml
[mcp_servers.flutter-skill] command = "npx" args = ["-y", "flutter-skill"]
Gemini CLI에 flutter-skill MCP 설치
$ gemini mcp add -s user flutter-skill npx -- -y flutter-skillGemini settings.json에 flutter-skill MCP 설치
~/.gemini/settings.json
{
"mcpServers": {
"flutter-skill": {
"command": "npx",
"args": [
"-y",
"flutter-skill"
]
}
}
}자동 검사
2026년 11월 14일에 실행됨전체 보고서 →자동 검사를 통과했습니다. 이는 보안 감사가 아닙니다. 코드를 직접 읽고 서버에 필요한 최소한의 권한만 부여하십시오. 검사 방식
- 소스 저장소 연결 및 접근 가능ai-dashboad/flutter-skill is public and reachablegithub.com/ai-dashboad/flutter-skill ↗통과
- 패키지가 이 저장소를 소스로 명시flutter-skill names ai-dashboad/flutter-skill as sourcenpmjs.com/package/flutter-skill ↗통과
- 빌드 출처(Sigstore 증명)No build attestation published for this release검사되지 않음
- 오픈 소스 라이선스 (OSI)MITspdx.org/licenses/MIT.html ↗통과
- 최근 1년 내 커밋 또는 릴리스Last commit or release 2026-09-01github.com/ai-dashboad/flutter-skill ↗통과
- OpenSSF Scorecard검사되지 않음
- 패키지 이름이 인기 패키지와 유사하지 않음No popular package one edit away from flutter-skill통과
- 패키지 및 해당 종속성에 중요한 보안 권고 없음the package checked; no critical advisoryosv.dev ↗통과
- 모델에 대한 숨겨진 지침 없음0 tool descriptions and the README read; no override phrases, credential paths or hidden characters통과
- 파괴적인 도구에 라벨 표시됨검사되지 않음
- 원격 엔드포인트가 MCP 핸드셰이크에 응답검사되지 않음
README
GitHub에서 보기30-Second Demo
https://github.com/user-attachments/assets/d4617c73-043f-424c-9a9a-1a61d4c2d3c6
One prompt. 28 AI-driven actions. Zero test code. The AI explores a TikTok clone, navigates tabs, scrolls feeds, tests search, fills forms — all autonomously.
Why This Exists
Writing E2E tests is painful. Maintaining them is worse. flutter-skill takes a different approach:
- 🔌 Connects any AI agent (Claude, Cursor, Windsurf, Copilot, OpenClaw) directly to your running app via MCP
- 👀 The agent sees your screen — taps buttons, types text, scrolls, navigates — like a human tester who never sleeps
- ✅ Zero test code — no Page Objects, no XPath, no brittle selectors. Just plain English
- ⚡ Zero config — 2 lines of code, works on all 10 platforms
You: "Test the checkout flow with an empty cart, then add 3 items and complete purchase"
Your AI agent handles the rest — screenshots, taps, text entry, assertions, navigation.
No Page Objects. No XPath. No brittle selectors. Just plain English.
Quick Start
1. Install (30 seconds)
npm install -g flutter-skill
2. Add to your AI (copy-paste into MCP config)
{
"mcpServers": {
"flutter-skill": {
"command": "flutter-skill",
"args": ["server"]
}
}
}
Works with Claude Desktop, Cursor, Windsurf, Copilot, Cline, OpenClaw — any MCP-compatible agent.
3. Add to your app (2 lines for Flutter)
import 'package:flutter_skill/flutter_skill.dart';
void main() {
if (kDebugMode) FlutterSkillBinding.ensureInitialized();
runApp(MyApp());
}
4. Test — just talk to your AI:
"Launch my app, explore every screen, and report any bugs"
That's it. Zero configuration. Zero test code. Works in under 60 seconds.
| Method | Command |
|---|---|
| npm | npm install -g flutter-skill |
| Homebrew | brew install ai-dashboad/flutter-skill/flutter-skill |
| Scoop | scoop install flutter-skill |
| Docker | docker pull ghcr.io/ai-dashboad/flutter-skill |
| pub.dev | dart pub global activate flutter_skill |
| VSCode | Extensions → "Flutter Skill" |
| JetBrains | Plugins → "Flutter Skill" |
| Agent Skill | npx skills add ai-dashboad/flutter-skill |
| Zero-config | flutter-skill init (auto-detects & patches your app) |
Use with AI Platforms
MCP Server Mode (IDE Integration)
Works with any MCP-compatible AI tool. One config line:
{
"mcpServers": {
"flutter-skill": {
"command": "flutter-skill",
"args": ["server"]
}
}
}
| Platform | Config File | Status |
|---|---|---|
| Cursor | .cursor/mcp.json |
✅ |
| Claude Desktop | claude_desktop_config.json |
✅ |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
✅ |
| VSCode Copilot | .vscode/mcp.json |
✅ |
| Cline | VSCode Settings → Cline → MCP | ✅ |
| OpenClaw | Skill or MCP config | ✅ |
| Continue.dev | .continue/config.json |
✅ |
HTTP Serve Mode (CLI & Automation)
For standalone browser automation, CI/CD pipelines, or remote access:
# Start server
flutter-skill serve https://your-app.com
# Use CLI client commands
flutter-skill nav https://google.com
flutter-skill snap # Accessibility tree (99% fewer tokens)
flutter-skill screenshot /tmp/ss.jpg
flutter-skill tap "Login"
flutter-skill type "hello@example.com"
flutter-skill eval "document.title"
flutter-skill tools # List all available tools
| Command | Description |
|---|---|
nav <url> |
Navigate to URL |
snap |
Accessibility tree snapshot |
screenshot [path] |
Take screenshot |
tap <text|ref|x y> |
Tap element |
type <text> |
Type via keyboard |
key <key> [mod] |
Press key |
eval <js> |
Execute JavaScript |
title |
Get page title |
text |
Get visible text |
hover <text> |
Hover element |
upload <sel> <file> |
Upload file |
tools |
List tools |
call <tool> [json] |
Call any tool |
Supports --port=N, --host=H flags and FS_PORT/FS_HOST env vars.
Two Modes Compared
server (MCP stdio) |
serve (HTTP) |
|
|---|---|---|
| Use case | IDE / AI agent integration | CLI / automation / CI/CD |
| Protocol | MCP (JSON-RPC over stdio) | HTTP REST |
| Tools | 253 (dynamic per page) | 246 (generic) |
| Browser | Auto-launches Chrome | Connects to existing Chrome |
| Best for | Cursor, Claude, VSCode | OpenClaw, scripts, pipelines |
Full CLI client reference: docs/CLI_CLIENT.md
10 Platforms, One Tool
Most testing tools work on 1-2 platforms. flutter-skill works on 10.
| Platform | SDK | Test Score |
|---|---|---|
| Flutter (iOS/Android/Web) | flutter_skill |
✅ 188/195 |
| React Native | sdks/react-native |
✅ 75/75 |
| Electron | sdks/electron |
✅ 75/75 |
| Tauri (Rust) | sdks/tauri |
✅ 75/75 |
| Android (Kotlin) | sdks/android |
✅ 74/75 |
| KMP Desktop | sdks/kmp |
✅ 75/75 |
| .NET MAUI | sdks/dotnet-maui |
✅ 75/75 |
| iOS (Swift/UIKit) | sdks/ios |
✅ 19/19 |
| Web (any website) | sdks/web |
✅ |
| Web CDP (zero-config) | No SDK needed | ✅ 141/156 |
Total: 656/664 tests passing (98.8%) — each platform tested against a complex social media app with 50+ elements.
⚡ Performance
Real benchmarks from automated test runs against a complex social media app:
| Operation | Web (CDP) | Electron | Android |
|---|---|---|---|
connect |
93 ms | 55 ms | 103 ms |
tap |
1 ms | 1 ms | 2 ms |
enter_text |
1 ms | 1 ms | 2 ms |
inspect |
3 ms | 12 ms | 10 ms |
snapshot |
2 ms | 8 ms | 29 ms |
screenshot |
31 ms | 80 ms | 88 ms |
eval |
1 ms | — | — |
Token efficiency: snapshot() returns a structured element tree instead of an image — 87–99% fewer tokens than sending screenshots to your AI agent.
How fast is that? A tap takes 1–2 ms end-to-end. Browser automation tools like Playwright and Selenium typically take 50–100 ms for the same operation. That's 50–100× faster, because flutter-skill talks directly to the app runtime instead of going through WebDriver or CDP indirection.
Heavy DOM Sites (Real-World)
Tested 15 MCP tools against production websites — 75/75 passed, zero timeouts:
| Site | Tools | Total Time | snapshot |
screenshot |
count_elements |
|---|---|---|---|---|---|
| YouTube | 15/15 ✅ | 6.9s | 43 ms | 30 ms | 4 ms |
| Amazon | 15/15 ✅ | 14.2s | 1 ms | 5 ms | 2 ms |
| 15/15 ✅ | 17.9s | 6 ms | 32 ms | 51 ms | |
| Hacker News | 15/15 ✅ | 4.8s | 53 ms | 188 ms | 1 ms |
| Wikipedia | 15/15 ✅ | 7.8s | 15 ms | 336 ms | 1 ms |
Total time includes page load. Tool execution is consistently sub-100ms even on heavy DOM sites.
Why Not Playwright / Appium / Detox?
| flutter-skill | Playwright MCP | Appium | Detox | |
|---|---|---|---|---|
| MCP tools | 253 | ~33 | ❌ | ❌ |
| Platforms | 10 | 1 (web) | Mobile | React Native |
| Setup time | 30 sec | Minutes | Hours | Hours |
| Test code needed | ❌ None | ✅ Yes | ✅ Yes | ✅ Yes |
| AI-native (MCP) | ✅ | ✅ | ❌ | ❌ |
| Self-healing tests | ✅ | ❌ | ❌ | ❌ |
| Monkey/fuzz testing | ✅ | ❌ | ❌ | ❌ |
| Visual regression | ✅ | ❌ | ❌ | ❌ |
| Network mock/replay | ✅ | ❌ | ❌ | ❌ |
| API + UI testing | ✅ | ❌ | ❌ | ❌ |
| Multi-device sync | ✅ | ❌ | Partial | ❌ |
| Accessibility audit | ✅ | ❌ | ❌ | ❌ |
| i18n testing | ✅ | ❌ | ❌ | ❌ |
| Performance monitoring | ✅ | ❌ | ❌ | ❌ |
| Natural language | ✅ | ❌ | ❌ | ❌ |
| Flutter support | ✅ Native | Partial | Partial | ❌ |
| Desktop apps | ✅ | ✅ | ❌ | ❌ |
| AI page understanding | ✅ AX Tree | ❌ Screenshots | ❌ | ❌ | | Boundary/security test | ✅ 13 payloads | ❌ | ❌ | ❌ | | Batch actions | ✅ 5+/call | 1/call | 1/call | 1/call |
flutter-skill is the only AI-native E2E testing tool that works across mobile, web, and desktop — with 7× more tools than the nearest competitor.
CLI Commands
# 🤖 AI autonomous exploration — finds bugs automatically
flutter-skill explore https://my-app.com --depth=3
# 🐒 Monkey/fuzz testing — random actions, crash detection
flutter-skill monkey https://my-app.com --actions=100 --seed=42
# 🚀 Parallel multi-platform testing
flutter-skill test --url https://my-app.com --platforms web,electron,android
# 🌐 Zero-config WebMCP server — any website becomes testable
flutter-skill serve https://my-app.com
🧠 AI-Native: 95% Fewer Tokens
Most AI testing tools send screenshots to the LLM — each one costs ~4,000 tokens.
flutter-skill uses Chrome's Accessibility Tree to give your AI a compact semantic summary of any page:
// page_summary → ~200 tokens (vs ~4,000 for a screenshot)
{
"title": "Shopping Cart",
"nav": ["Home", "Products", "Cart", "Account"],
"forms": [{"input:Coupon Code": "text"}],
"buttons": ["Apply", "Checkout", "Continue Shopping"],
"features": {"search": true, "pagination": true},
"links": 47, "inputs": 3
}
Then batch multiple actions in one call:
// explore_actions → 5 actions per call (vs 5 separate tool calls)
{"actions": [
{"type": "fill", "target": "input:Coupon Code", "value": "SAVE20"},
{"type": "tap", "target": "button:Apply"},
{"type": "tap", "target": "button:Checkout"},
{"type": "fill", "target": "input:Email", "value": "test@example.com"},
{"type": "tap", "target": "button:Continue"}
]}
Result: Your AI agent tests faster, costs less, and understands pages better than screenshot-based tools.
| flutter-skill | Screenshot-based tools | |
|---|---|---|
| Tokens per page | ~200 | ~4,000 |
| Actions per call | 5+ | 1 |
| Understands semantics | ✅ roles, names, state | ❌ pixels only |
| Works with Shadow DOM | ✅ | ❌ |
What It Can Do
👀 See
screenshot— capture the screeninspect_interactive— all tappable/typeable elements with semantic refsfind_element/wait_for_elementget_elements— full element tree
👆 Interact
tap/long_press/swipe/dragenter_text/set_text/clear_textscroll— all directionsgo_back/press_key
🔍 Inspect (v0.8.0)
- Semantic refs:
button:Login,input:Email - Stable across UI changes
tap(ref: "button:Submit")- 7 roles: button, input, toggle, slider, select, link, item
🚀 Control
launch_app— launch with flavorshot_reload/hot_restartget_logs/get_errorsscan_and_connect— auto-find apps
AI Explore: page_summary, explore_actions, boundary_test, explore_report
Launch & Connect: launch_app, scan_and_connect, connect_cdp, hot_reload, hot_restart, list_sessions, switch_session, close_session, disconnect, stop_app
Screen: screenshot, screenshot_region, screenshot_element, native_screenshot, inspect, inspect_interactive, snapshot, get_widget_tree, find_by_type, get_text_content, get_visible_text
Interaction: tap, double_tap, long_press, enter_text, set_text, clear_text, swipe, scroll_to, drag, go_back, press_key, type_text, hover, fill, select_option, set_checkbox, focus, blur, native_tap, native_input_text, native_swipe
Smart Testing: smart_tap, smart_enter_text, smart_assert (self-healing with fuzzy match)
Assertions: assert_text, assert_visible, assert_not_visible, assert_element_count, assert_batch, wait_for_element, wait_for_gone, wait_for_idle, wait_for_stable, wait_for_url, wait_for_text, wait_for_element_count
Visual Regression: visual_baseline_save, visual_baseline_compare, visual_baseline_update, visual_regression_report, visual_verify, visual_diff, compare_screenshot
Network Mock: mock_api, mock_clear, record_network, replay_network, intercept_requests, clear_interceptions, block_urls, http_request
API Testing: api_request, api_assert
Coverage & Reliability: coverage_start, coverage_stop, coverage_report, coverage_gaps, retry_on_fail, stability_check
Data-Driven: test_with_data, generate_test_data
Multi-Device: multi_connect, multi_action, multi_compare, multi_disconnect, parallel_snapshot, parallel_tap
Accessibility: accessibility_audit, a11y_full_audit, a11y_tab_order, a11y_color_contrast, a11y_screen_reader
i18n: set_locale, verify_translations, i18n_snapshot
Performance: perf_start, perf_stop, perf_report, get_performance, get_frame_stats, get_memory_stats
Session: save_session, restore_session, session_diff
Recording & Export: record_start, record_stop, record_export (Playwright, Cypress, XCUITest, Espresso, Detox, Maestro, +5 more), video_start, video_stop
Auth: auth_inject_session, auth_biometric, auth_otp, auth_deeplink
CDP Browser: navigate, reload, go_forward, get_title, get_page_source, eval, get_tabs, new_tab, switch_tab, close_tab, get_cookies, set_cookie, clear_cookies, get_local_storage, set_local_storage, clear_local_storage, generate_pdf, set_viewport, emulate_device, throttle_network, go_offline, set_geolocation, set_timezone, set_color_scheme
Debug: get_logs, get_errors, get_console_messages, get_network_requests, diagnose, diagnose_project, reset_app
Platform Setup
dependencies:
flutter_skill: ^0.9.37
import 'package:flutter_skill/flutter_skill.dart';
void main() {
if (kDebugMode) FlutterSkillBinding.ensureInitialized();
runApp(MyApp());
}
npm install flutter-skill-react-native
import FlutterSkill from 'flutter-skill-react-native';
FlutterSkill.start();
npm install flutter-skill-electron
const { FlutterSkillBridge } = require('flutter-skill-electron');
FlutterSkillBridge.start(mainWindow);
// Swift Package Manager: FlutterSkillSDK
import FlutterSkill
FlutterSkillBridge.shared.start()
Text("Hello").flutterSkillId("greeting")
implementation("com.flutterskill:flutter-skill:0.8.0")
FlutterSkillBridge.start(this)
[dependencies]
flutter-skill-tauri = "0.8.0"
Add Gradle dependency — see sdks/kmp for details.
Add NuGet package — see sdks/dotnet-maui for details.
Example Prompts
Just tell your AI what to test:
| Prompt | What happens |
|---|---|
| "Test login with wrong password" | Screenshots → enters creds → taps login → verifies error |
| "Explore every screen and report bugs" | Systematically navigates all screens, tests all elements |
| "Fill registration with edge cases" | Tests emoji 🌍, long strings, empty fields, special chars |
| "Compare checkout flow on iOS and Android" | Runs same test on both platforms, compares screenshots |
| "Take screenshots of all 5 tabs" | Taps each tab, captures state |
Contributing
See CONTRIBUTING.md for guidelines.
git clone https://github.com/ai-dashboad/flutter-skill
cd flutter-skill
dart pub get
dart run bin/flutter_skill.dart server # Start MCP server
Links
| 📦 pub.dev | 🧩 VSCode |
| 📦 npm | 🧩 JetBrains |
| 🍺 Homebrew | 📖 Docs |
| 🤖 Agent Skill | 📋 Changelog |
flutter-skill MCP 질문
flutter-skill MCP를 Claude Code에 추가하려면 어떻게 하나요?
터미널에서 claude mcp add flutter-skill -- npx -y flutter-skill를 실행합니다. 모든 프로젝트에서 사용할 수 있게 하려면 --scope user를 추가합니다.
flutter-skill MCP를 Cursor에 추가하려면 어떻게 하나요?
~/.cursor/mcp.json를 열고 mcpServers 아래에 서버를 추가합니다. 위 설치 섹션에 표시된 Cursor 구성을 사용합니다.
flutter-skill을 어떻게 설치하나요?
npm(또는 Homebrew, Scoop, Docker, pub.dev)을 사용하여 전역으로 설치한 다음, MCP 설정에 추가하고 앱에 바인딩을 등록하세요.
flutter-skill로 어떤 플랫폼을 테스트할 수 있나요?
Flutter, React Native, Electron, Tauri, Android, KMP Desktop, .NET MAUI, iOS 및 모든 웹사이트를 포함하여 총 10개의 플랫폼을 지원합니다.
flutter-skill이 Claude, Cursor 또는 Copilot과 연동되나요?
네, MCP server로 동작하며 Claude Desktop, Cursor, Windsurf, VS Code Copilot 등 MCP 호환 에이전트와 함께 사용할 수 있습니다.
flutter-skill을 사용하려면 테스트 코드를 작성해야 하나요?
아니요, 테스트 내용을 평문 영어로 설명하면 AI 에이전트가 스크린샷 캡처, 탭, 타이핑, 네비게이션을 대신 수행합니다.
스크린샷 기반 도구와 비교했을 때 flutter-skill의 토큰 효율성은 어떤가요?
이미지 대신 구조화된 accessibility tree 스냅샷을 전송하므로 페이지당 토큰 사용량이 훨씬 적습니다.