flutter-dev-mcp

flutter-dev-mcp

An MCP server that wraps Flutter CLI tools into structured, agent-friendly tools for AI coding agents, with features like two-phase test results, output capping, and managed app lifecycle.

Category
访问服务器

README

flutter-dev-mcp

An MCP (Model Context Protocol) server that gives AI coding agents first-class Flutter development tools.

Why?

Flutter CLI tools are designed for humans, not agents. flutter test dumps hundreds of lines of output that overwhelm context windows. flutter run requires interactive terminal access for hot reload. flutter analyze produces unstructured text. This MCP server wraps the Flutter CLI into structured, agent-friendly tools with sensible output limits.

Key design decisions:

  • Test results are two-phase: flutter_test returns a compact summary of failures. flutter_get_result fetches full error details for specific tests. This prevents a single test run from blowing the context window.
  • All outputs are capped at 24KB to stay within typical tool response limits.
  • flutter run is managed: The server holds the process, exposes hot reload/restart/logs/kill as separate tools, so the agent doesn't need terminal access.
  • Inputs are sanitized: All commands use array-based process spawning (no shell). Project paths are normalized and validated. Package names and device IDs are checked for flag injection.

Tools at a glance

Tool Parameters Description
flutter_test project_dir, [test_path], [test_name], [extra_args] Run tests and return a compact summary of failures only. Use flutter_get_result to drill into specific failures.
flutter_get_result test_run_id, test_ids Get full error details for specific test IDs from a previous flutter_test run. Output capped at 24KB.
flutter_run project_dir, [device], [is_debug], [dont_detach], [extra_args] Start a Flutter app on device (e.g. macos, chrome, emulator ID) in debug or release mode. By default detaches after the app starts and returns a run_id.
flutter_hot_reload run_id Hot reload a running app.
flutter_hot_restart run_id Hot restart a running app.
flutter_kill run_id Kill a running app. Graceful shutdown, force-kills after 5s.
flutter_logs run_id Get logs from a running app. Returns the most recent output, capped at 24KB.
flutter_analyze project_dir Run static analysis. Returns structured issues with severity, file, line, column, and rule name.
flutter_devices [wireless] List available devices (simulators, emulators, physical). Skips wireless scan by default.
flutter_clean project_dir Delete build artifacts. Useful when builds get into a bad state.
flutter_pub_get project_dir Resolve and download dependencies.
flutter_pub_add project_dir, packages, [dev] Add one or more packages. Supports dev dependencies.
flutter_gen_l10n project_dir Generate localization files from ARB files.
flutter_build project_dir, target, [debug], [extra_args] Build the app for a target platform (apk, ios, web, macos, etc.).
flutter_build_runner project_dir, [delete_conflicting] Run build_runner for code generation (freezed, json_serializable, drift, etc.).

Parameters in [brackets] are optional.

Install

Requires Node.js 18+ and Flutter SDK on your PATH.

npm install -g flutter-dev-mcp

Or run directly with npx (recommended):

npx -y flutter-dev-mcp

Options

--limit-tools  Only expose tools that provide significant benefit over
               direct CLI usage (testing, app lifecycle, logs). Omits
               analyze, devices, clean, pub get/add, gen-l10n, build,
               and build_runner, which agents can run via shell.

Configuration

Claude Code (CLI)

Easy: claude mcp add flutter-dev -- npx -y flutter-dev-mcp

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "flutter-dev": {
      "command": "npx",
      "args": ["-y", "flutter-dev-mcp"]
    }
  }
}

Codex / Other agents

Any agent that supports MCP can use this server. Point it at the stdio transport:

npx -y flutter-dev-mcp

The server communicates over stdin/stdout using the MCP JSON-RPC protocol.

Tools

Testing

flutter_test

Run tests and get a summary of failures.

Parameter Type Required Description
project_dir string yes Path to the Flutter project
test_path string no Specific test file or directory
test_name string no Filter by test name (plain string match)
extra_args string[] no Additional flags (e.g. ["--coverage", "--dart-define=KEY=VALUE"])

Returns a test_run_id and an array of failed tests with short error excerpts. Pass the test_run_id to flutter_get_result for full details.

flutter_get_result

Get full error output for specific test IDs from a previous run.

Parameter Type Required Description
test_run_id number yes From a previous flutter_test call
test_ids number[] yes Test IDs to get details for

Output is capped at 24KB total. If a single test exceeds that, its error is truncated. If multiple tests would exceed it, only tests that fit are returned.

App lifecycle

flutter_run

Start a Flutter app and get a run_id for subsequent commands.

Parameter Type Required Default Description
project_dir string yes Path to the Flutter project
device string no "" Device ID (e.g. chrome, macos, emulator ID)
is_debug boolean no true Debug mode (true) or release mode (false)
dont_detach boolean no false Wait for app to exit instead of returning after start
extra_args string[] no [] Additional flags (e.g. ["--flavor=dev", "--dart-define=KEY=VALUE"])

flutter_hot_reload

Trigger a hot reload on a running app.

Parameter Type Required Description
run_id number yes From a previous flutter_run call

flutter_hot_restart

Trigger a hot restart on a running app.

Parameter Type Required Description
run_id number yes From a previous flutter_run call

flutter_kill

Kill a running app. Sends q for graceful shutdown, force-kills after 5s.

Parameter Type Required Description
run_id number yes From a previous flutter_run call

flutter_logs

Get logs from a running app. Returns the most recent output, capped at 24KB.

Parameter Type Required Description
run_id number yes From a previous flutter_run call

Analysis

flutter_analyze

Run static analysis. Returns structured issues with severity, file location, and lint rule name.

Parameter Type Required Description
project_dir string yes Path to the Flutter project

flutter_devices

List available devices.

Parameter Type Required Default Description
wireless boolean no false Include wireless devices (slower)

Build

flutter_build

Build the app for a target platform. Can take a long time for release builds.

Parameter Type Required Default Description
project_dir string yes Path to the Flutter project
target string yes Build target: apk, appbundle, ios, ipa, web, macos, windows, linux, aar, bundle, ios-framework, macos-framework
debug boolean no true Debug mode (true) or release mode (false)
extra_args string[] no [] Additional flags (e.g. ["--simulator", "--flavor=dev"])

Dependencies & codegen

flutter_pub_get

Resolve and download dependencies.

Parameter Type Required Description
project_dir string yes Path to the Flutter project

flutter_pub_add

Add packages to the project.

Parameter Type Required Default Description
project_dir string yes Path to the Flutter project
packages string[] yes Package names (e.g. ["http", "provider"])
dev boolean no false Add as dev dependency

flutter_clean

Delete build artifacts. Useful when builds get into a bad state.

Parameter Type Required Description
project_dir string yes Path to the Flutter project

flutter_gen_l10n

Generate localization files from ARB files.

Parameter Type Required Description
project_dir string yes Path to the Flutter project

flutter_build_runner

Run dart run build_runner build for code generation (freezed, json_serializable, drift, etc.).

Parameter Type Required Default Description
project_dir string yes Path to the Flutter project
delete_conflicting boolean no true Delete conflicting outputs before building

Building from source

git clone <repo-url>
cd flutter-dev-mcp
npm install
npm run build
node dist/index.js

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选