mcpField
F
FlutterSDK Docs

FlutterSDK Docs

Remote
fluttersdk.comDeveloper Tools3Updated today

About FlutterSDK Docs

FlutterSDK Docs provides documentation search and AI coding skills for the FlutterSDK ecosystem, including Wind UI and Magic Framework. It is for coding agents and developers using these frameworks, with skills distributed across multiple coding tools and a public HTTP MCP service.

Key features

  1. 01Five FlutterSDK skills
  2. 02Wind UI and Magic documentation
  3. 03Distribution to coding tools
  4. 04Public documentation search

Use cases

  1. 01Look up FlutterSDK framework usage
  2. 02Give coding agents framework conventions
  3. 03Search framework docs while coding

MCP support

Its Streamable HTTP MCP server exposes a `search-docs` tool without authentication. Clients limited to stdio can use the `npx @fluttersdk/mcp` bridge.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://mcp.fluttersdk.com/
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Packages
npm@fluttersdk/mcp1.1.3

Tools

1
ParameterTypeRequiredDescription
queriesarrayNoList of search queries. Pass multiple when terminology is ambiguous (e.g. ['toggle', 'switch']). Required.
versionstringNoOptional version pin, applied to both the pages found and the content served. Default: latest per package. On a miss the series chain falls back and the response header names what was requested.
packagesarrayNoOptional package filter. Slugs (artisan, dusk, magic, telescope, wind) or composer-style (fluttersdk_wind). Default: all 5.
token_limitintegerNoToken budget for response. Default 3000, max 100000. Approx chars = tokens × 4.

Details

From the project README

FlutterSDK AI Logo

FlutterSDK AI

AI coding skills for the FlutterSDK ecosystem.
Teach your AI assistant Wind UI, Magic Framework, and more, across every major coding tool.

Registry License: MIT GitHub stars

Registry · Issues · Quick Start


Why?

AI coding assistants are powerful, but they don't know your frameworks. They hallucinate widget names, guess at API patterns, and miss conventions.

FlutterSDK AI fixes this. It gives your AI structured knowledge about Wind UI's className system, Magic Framework's Laravel-inspired architecture, and more, so it generates correct code on the first try.

One install, every tool:

ToolIntegration
OpenCodeURL-based skill registry
Claude CodePlugin + Marketplace
CursorRules (.mdc)
Gemini CLICommand templates (.toml)
VS Code CopilotInstructions (.md)
Codex CLIMCP config (~/.codex/config.toml)
Antigravity CLIMCP config (~/.gemini/config/mcp_config.json)
ClineMCP config (.mcp.json)
Roo CodeMCP config (.mcp.json)

Skills

Five skills are available. See docs/SKILLS.md for full descriptions.

SkillPackageUpstreamWhat it covers
wind-uifluttersdk_windfluttersdk/windW-prefix widgets, the className parser pipeline, WindRecipe variants, dark mode, responsive breakpoints, design culture
magic-frameworkmagicfluttersdk/magicIoC container, 18 facades, Eloquent ORM, GoRouter routing, forms, auth, testing, 5 ecosystem plugins
fluttersdk-duskfluttersdk_duskfluttersdk/duskE2E driver: dusk_* MCP tools and matching CLI commands to snap, tap, type, and screenshot a running Flutter app
fluttersdk-telescopefluttersdk_telescopefluttersdk/telescopeRuntime inspector: 9 ring buffers for HTTP, logs, exceptions, events, gates, DB queries
fluttersdk-artisanfluttersdk_artisanfluttersdk/artisanCLI framework + stdio MCP server: substrate artisan_* tools, builtin commands, artisan_tinker

Each skill mirrors its upstream package's own skills/ directory and arrives here when that package publishes a release. The exact synced version is the version: field in skills/<name>/SKILL.md, and skills/index.json is the machine-readable index; neither is duplicated in this table, so neither can drift from it.

MCP Server

mcp.fluttersdk.com exposes a single search-docs tool via Streamable HTTP. Public, no auth required. Full client config snippets (Claude Code, Cursor, OpenCode, Codex CLI, Gemini CLI, VS Code Copilot, and more) are in docs/MCP.md.

For clients that only support stdio (Claude Desktop, older Codex CLI):

npx @fluttersdk/mcp

The npx bridge proxies stdio to the upstream HTTP server transparently.

Quick Start

Option 1: npx skills (Recommended)

The fastest way. Works with 40+ AI coding agents.

# Install all FlutterSDK skills to all detected agents
npx skills add fluttersdk/ai

# Install a specific skill
npx skills add fluttersdk/ai --skill wind-ui

# Install globally (available across all projects)
npx skills add fluttersdk/ai -g
Option 2: Universal Installer
git clone https://github.com/fluttersdk/ai.git && cd ai

# Preview what would be installed
bash scripts/install.sh --dry-run --global

# Install globally for all detected tools
bash scripts/install.sh --global

# Install globally and register the MCP server for all detected tools
bash scripts/install.sh --global --with-mcp

# Or install into your current Flutter project
bash scripts/install.sh --project
Option 3: Tool-Specific
OpenCode

Add the registry URL to your opencode.json:

{
    "skills": {
        "urls": ["https://fluttersdk.github.io/ai/"]
    }
}

Skills are fetched automatically and cached locally.

Claude Code
/plugin marketplace add https://raw.githubusercontent.com/fluttersdk/ai/main/.claude-plugin/marketplace.json
/plugin install fluttersdk@fluttersdk-marketplace

The plugin auto-registers:

  • 5 skills (/fluttersdk:wind-ui, /fluttersdk:magic-framework, /fluttersdk:fluttersdk-dusk, /fluttersdk:fluttersdk-telescope, /fluttersdk:fluttersdk-artisan)
  • /fluttersdk:flutter-review, a review that runs in a subagent as an independent second pair of eyes and then fixes what it finds, plus /fluttersdk:flutter-test
  • fluttersdk:flutter-reviewer and fluttersdk:flutter-reviewer-deep subagents, invokable directly with @
  • fluttersdk MCP server (HTTP transport to mcp.fluttersdk.com; no manual .mcp.json edit required)
Cursor
mkdir -p .cursor/rules
cp tool-templates/cursor/fluttersdk.mdc .cursor/rules/

Activates automatically for all .dart files.

Gemini CLI
mkdir -p ~/.gemini/commands
cp tool-templates/gemini/*.toml ~/.gemini/commands/

Use with /flutter-review and /flutter-test commands.

VS Code Copilot
mkdir -p .github
cp .github/copilot-instructions.md .github/

How It Works

skills/index.json              ← OpenCode fetches this from GitHub Pages
  → wind-ui/SKILL.md           ← Skill definition + className rules
  → wind-ui/references/*.md    ← On-demand widget docs, layout patterns, tokens
  → magic-framework/SKILL.md   ← Skill definition + framework laws
  → magic-framework/refs/*.md  ← Controllers, ORM, facades, routing, auth, ...

OpenCode: Fetches fluttersdk.github.io/ai/index.json, downloads listed skill files, caches locally.

Claude Code: Plugin manifest (.claude-plugin/plugin.json) makes the repo installable. Skills in skills/ are auto-discovered and namespaced under /fluttersdk:.

Deployment: Push to main → GitHub Actions deploys skills/ to GitHub Pages automatically.

Contributing

Adding a New Skill
  1. Create skills/<skill-name>/SKILL.md with YAML frontmatter (name, description)
  2. Add reference files in skills/<skill-name>/references/
  3. Update skills/index.json: list every file in the files array
  4. Bump version in .claude-plugin/plugin.json and .claude-plugin/marketplace.json
  5. Push to main

See CLAUDE.md for the full specification.

Important: Every file must be listed in index.json: OpenCode only downloads listed files. Forgetting a file means users get incomplete skills.

License

MIT. See LICENSE for details.


Built with care by FlutterSDK
If this saves you time, give it a star; it helps others discover it.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.