community-ff-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@community-ff-mcp List my FlutterFlow projects"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
community-ff-mcp
Disclaimer: This is an unofficial, community-built project. It is not affiliated with, endorsed by, or sponsored by FlutterFlow. "FlutterFlow" is a trademark of FlutterFlow Inc. This project uses the FlutterFlow public API.
MCP server for the FlutterFlow Project API. Enables AI-assisted FlutterFlow development through Claude and other MCP-compatible clients.
What This MCP Does Best
This MCP excels at reading, exploring, and understanding your FlutterFlow projects. Here's what it's great at:
Project exploration — Browse your projects, pages, components, and file structure
Understanding functionality — Get clear summaries of what a page or component does, including its widget tree, actions, parameters, and state
Tracing usage — Find everywhere a component is used, or every navigation action that leads to a page
Inspecting configuration — View theme settings, API endpoints, data models, custom code, app state, integrations, and more
Searching — Search project files by keyword, prefix, or regex to find exactly what you need
YAML reference — Built-in documentation for FlutterFlow's YAML schema, so your AI assistant can understand and generate valid YAML
Guided editing — A workflow guide (
get_editing_guide) that walks through the correct steps before making any YAML changes
In short: Think of it as giving your AI assistant full read access to your FlutterFlow project, plus careful write access when needed.
Related MCP server: Claude MCP Server Integration
Quick Start
1. Get Your FlutterFlow API Token
You need a FlutterFlow API token to authenticate. Here's how to get one:
Open FlutterFlow and log in
Click your profile picture (bottom-left corner)
Go to Account Settings
Scroll to the API Token section
Click Copy to copy your token
Note: The API token requires a paid FlutterFlow subscription (Standard plan or above). Free-tier accounts do not have API access.
2. Add to Your MCP Client
Choose your AI client below and follow the setup instructions.
Run this command to add the MCP server:
claude mcp add flutterflow -e FLUTTERFLOW_API_TOKEN=your_token_here -- npx -y community-ff-mcpReplace your_token_here with the token you copied in step 1.
This adds the server to your project's .claude/settings.json. You can also manually edit the config:
{
"mcpServers": {
"flutterflow": {
"command": "npx",
"args": ["-y", "community-ff-mcp"],
"env": {
"FLUTTERFLOW_API_TOKEN": "your_token_here"
}
}
}
}Open Settings > Developer > Edit Config and add:
{
"mcpServers": {
"flutterflow": {
"command": "npx",
"args": ["-y", "community-ff-mcp"],
"env": {
"FLUTTERFLOW_API_TOKEN": "your_token_here"
}
}
}
}Replace your_token_here with your token. Restart Claude Desktop after saving.
Open Settings > MCP and add a new server with this configuration:
{
"mcpServers": {
"flutterflow": {
"command": "npx",
"args": ["-y", "community-ff-mcp"],
"env": {
"FLUTTERFLOW_API_TOKEN": "your_token_here"
}
}
}
}Add the following to your MCP configuration file (check your client's docs for the exact location):
{
"mcpServers": {
"flutterflow": {
"command": "npx",
"args": ["-y", "community-ff-mcp"],
"env": {
"FLUTTERFLOW_API_TOKEN": "your_token_here"
}
}
}
}3. Start Building
Ask your AI assistant to list your FlutterFlow projects, inspect pages, or explore your app — it handles the rest.
Example prompts to get started:
"List my FlutterFlow projects"
"Show me all the pages in project X"
"What does the HomePage do? Walk me through its widget tree"
"Where is the PaywallCard component used?"
"What API endpoints are configured?"
"Show me the theme colors"
How It Works Under the Hood
This MCP is designed to make AI agents as effective as possible when working with FlutterFlow. Here's what makes it work well:
Smart Local Caching
Instead of hitting the FlutterFlow API for every question, the MCP downloads your entire project once (sync_project) and caches it locally. After that, most tools read from the cache — making them instant and free of API rate limits. If you've made changes in FlutterFlow, just re-sync to refresh the cache.
Built-in Documentation
AI models don't natively understand FlutterFlow's internal structure. This MCP ships with 21 reference documents covering every widget type, action, variable, theme setting, and editing pattern. The AI can look up exactly how a Button, Column, or Navigation action is structured — so it doesn't have to guess.
Guided Editing Workflow
Before making any change, the AI can call get_editing_guide with a plain-English description of what it wants to do (e.g., "change the button color on the login page"). The MCP returns the exact steps, relevant documentation, and rules to follow — reducing the chance of mistakes.
25 Purpose-Built Tools
Rather than one generic "read project" tool, the MCP provides 25 specialized tools — each designed for a specific task. Want to know where a component is used? There's a dedicated tool for that. Want to see all navigation actions pointing to a page? There's one for that too. This means the AI always picks the right tool for the job instead of downloading everything and searching through it.
Limitations and Cautions
Be Careful with Edits
This MCP can make changes to your FlutterFlow project, but edits should be treated with care:
Always review before pushing — Ask your AI assistant to show you exactly what it plans to change before it pushes anything. Make sure you understand the change.
There's no undo button for API changes — Changes are applied immediately to your project. You can roll back using FlutterFlow's version history, but it's better to prevent a bad edit than to fix one.
Start with read-only tasks — Get comfortable asking questions about your project (page summaries, component usage, theme inspection) before asking the AI to modify anything.
Smaller edits are safer — Editing a single widget is safer than replacing an entire page. The AI knows to make targeted changes, but it's good to be aware.
What the FlutterFlow API Can't Do
These are limitations in FlutterFlow's own API. They affect any tool that connects to FlutterFlow, not just this MCP.
Custom code can be read but not edited — The AI can read your custom actions, custom functions, and custom widgets, but it cannot push code changes back to FlutterFlow. The API doesn't support it properly — attempting to push code corrupts the FlutterFlow editor. Instead, the AI will show you the modified code and you copy-paste it into the FlutterFlow code editor.
Unlocked
main.dartcan't be updated — If you've unlockedmain.dartfor raw Dart editing, the AI can read it but can't push code changes. Startup/shutdown action lists can still be modified.Platform config files are indirect — You can't directly edit
AndroidManifest.xml,Info.plist, orbuild.gradle. Instead, you configure hooks and settings in FlutterFlow, which generates those files for you. The AI can read and modify those hooks.Editing platform config files requires extra care — FlutterFlow groups all platform config files together internally. If the AI pushes a change to just one (e.g., Android Manifest), FlutterFlow may delete the others (e.g., ProGuard rules, Gradle config). The MCP handles this by always including all existing files in the same push, but it's something to be aware of.
Disabling conditional actions works differently — In the FlutterFlow editor, you can toggle a conditional action on/off with a switch. The API handles this slightly differently, so the AI uses a workaround that achieves the same result.
Validation isn't perfect — The validation step catches most errors, but occasionally something that passes validation can still fail when pushed. This is rare, but it's why reviewing changes before pushing is important.
Project list may be incomplete — The API may not return all projects you have access to, especially shared or team projects. If a project is missing, copy its ID directly from the FlutterFlow editor (click the project name in the top-left corner) and paste it into your AI prompt.
General Limitations
No live sync — The MCP works with a snapshot of your project. It doesn't watch for changes you make in the FlutterFlow editor in real time. If you've been editing in FlutterFlow, re-sync before asking the AI to make changes.
Very large pages — Some extremely large pages may fail to load. The AI can work around this by loading individual widgets instead of the whole page.
Cache can go stale — If you edit your project in FlutterFlow after syncing, the AI's local copy is outdated. Re-run the sync (the AI will tell you when it detects stale data).
For the full technical details on each API limitation, see
docs/flutterflow-api-limitations.md.
Tools
Discovery and Exploration
Tool | Description |
| List all FlutterFlow projects for your account |
| List YAML file keys in a project (supports prefix filter) |
| List all pages with human-readable names, scaffold IDs, and folders |
| Search file keys by keyword, prefix, or regex |
| Bulk download all project YAML to local cache for fast offline reads |
Reading and Understanding
Tool | Description |
| Fetch a page by its human-readable name |
| Download specific YAML files by file key |
| Quick page overview from cache — widget tree, actions, params, state |
| Quick component overview from cache — widget tree, actions, params |
| Find all pages and components where a given component is used |
| Find all actions that navigate to a given page |
Project Configuration
Tool | Description |
| Theme colors, typography, breakpoints, and widget defaults |
| App state variables, constants, and environment settings |
| API endpoint definitions — method, URL, variables, headers, response |
| Data structs, enums, Firestore collections, and Supabase tables |
| Custom actions, functions, widgets, AI agents, and app-action components |
| App Details, App Assets, Nav Bar & App Bar settings |
| Firebase, Languages, Platforms, Permissions, Dependencies |
| Authentication, Push Notifications, Mobile/Web Deployment |
| Stripe, Braintree, RevenueCat, Razorpay configuration |
| Supabase, SQLite, GitHub, Algolia, Google Maps, AdMob, and more |
Editing
Tool | Description |
| Get the recommended workflow and docs for an editing task — call this before modifying any YAML |
| Search/retrieve FlutterFlow YAML reference docs by topic or file |
| Validate YAML content before pushing — always call before |
| Push YAML changes to a project |
Resources
Resource | URI | Description |
Docs Index |
| List all available YAML reference documentation files |
Doc File |
| Read a specific YAML reference doc (e.g. |
Prompts
Prompt | Description |
| Generate a new page from a description |
| Modify an existing component |
| Summarize project structure |
| Efficient workflow guide for AI-assisted FlutterFlow development |
Built-in YAML Reference
This MCP ships with a comprehensive FlutterFlow YAML reference catalog that AI models can access at runtime:
get_yaml_docstool — Search by topic (e.g.get_yaml_docs(topic: "Button")) or browse the full indexff://docsresources — Direct access to all 21 reference docs covering widgets, actions, variables, theming, and editing workflows
See docs/ff-yaml/ for the full catalog.
AI Agent Skill
This MCP includes a skills.sh-compatible skill that teaches AI assistants how to use FlutterFlow MCP effectively. The skill provides tool orchestration workflows, critical YAML rules, and reference documentation for widgets, actions, data binding, theming, and more.
Install the Skill
npx skills add mohn93/ff-mcp --skill community-ff-mcpThis works with 18+ AI agents including Claude Code, Cursor, GitHub Copilot, Codex, Goose, Windsurf, and more. The skills CLI auto-detects your AI tool and installs the skill to the correct location.
What the Skill Teaches Your AI
Once installed, your AI assistant will know how to:
Discover — List projects, pages, components, and search project files
Read — Summarize pages/components, trace component usages, inspect navigations
Edit — Follow the correct read → guide → edit → validate → push workflow
Configure — Read theme, app state, API endpoints, data models, and integrations
Skill + MCP Setup
The skill and the MCP server are installed separately:
MCP server gives the AI the tools (see Quick Start above)
Skill teaches the AI how to use those tools effectively
If the skill detects the MCP tools aren't connected, it will guide the user through setup automatically.
Claude Code Skills
You can also copy Claude Code-specific skills from skills/ into your project's .claude/skills/ directory:
ff-yaml-dev.md— Core workflow: reading, editing, and creating FlutterFlow pages/componentsff-widget-patterns.md— Quick reference for common widget YAML patterns and snippets
Development
git clone https://github.com/mohn93/ff-mcp.git
cd ff-mcp
npm install
npm run build
npm run dev # watch modeRequirements
Node.js 18+
FlutterFlow API token (paid FlutterFlow subscription required)
License
MIT
Available Tools
25 toolsfind_component_usagesA
Find all pages and components where a given component is used, with parameter pass details. Cache-only, no API calls. Run sync_project first.
Inputs: projectId (required), componentId, componentName.
get_api_endpointsA
Get API endpoint definitions from local cache — method, URL, variables, headers, response fields. No API calls. Run sync_project first if not cached.
Inputs: projectId (required), name.
get_app_settingsA
Get App Settings — Authentication, Push Notifications, Mobile Deployment (version, build, stores), Web Deployment (SEO, title). Mirrors the FlutterFlow 'App Settings' section. Cache-based, no API calls. Run sync_project first.
Inputs: projectId (required).
get_app_stateA
Get app state variables, constants, and environment settings from local cache. No API calls. Run sync_project first if not cached.
Inputs: projectId (required).
get_component_summaryA
Get a readable summary of a FlutterFlow component from local cache — widget tree, actions, params. Nested component references are resolved to show [ComponentName] (ComponentId). No API calls. Run sync_project first if not cached.
Inputs: projectId (required), componentId, componentName.
get_custom_codeA
Get custom actions, functions, widgets, AI agents, app action components, and custom files from local cache — signatures, arguments, return types, and optionally Dart source code. No API calls. Run sync_project first if not cached.
Inputs: projectId (required), name, type, includeCode.
get_data_modelsA
Get data structs, enums, Firestore collections, and Supabase tables from local cache. No API calls. Run sync_project first if not cached.
Inputs: projectId (required), name, type.
get_editing_guideA
Get the recommended workflow and relevant documentation for a FlutterFlow editing task. Call this BEFORE modifying any YAML. Describe what you want to do (e.g. 'change button color', 'add a TextField to the login page', 'create a reusable header component') and receive the correct workflow steps, YAML schemas, and critical rules.
Inputs: task (required), projectId.
get_general_settingsA
Get General settings — App Details (name, package, initial page, routing), App Assets (icon, splash, error image), Nav Bar & App Bar. Mirrors the FlutterFlow 'General' settings section. Cache-based, no API calls. Run sync_project first.
Inputs: projectId (required).
get_in_app_purchasesA
Get In-App Purchases & Subscriptions settings — Stripe, Braintree, RevenueCat, Razorpay. Mirrors the FlutterFlow 'In App Purchases & Subscriptions' section. Cache-based, no API calls. Run sync_project first.
Inputs: projectId (required).
get_integrationsA
Get Integrations settings — Supabase, SQLite, GitHub, Algolia, Google Analytics, Google Maps, AdMob, Mux Livestream, OneSignal, Gemini. Mirrors the FlutterFlow 'Integrations' section. Cache-based, no API calls. Run sync_project first.
Inputs: projectId (required).
get_page_by_nameA
Fetch a FlutterFlow page by its human-readable name (e.g. 'Welcome', 'GoldPass'). Resolves the name to the correct scaffold ID and returns the full page YAML. Case-insensitive matching.
Inputs: pageName (required), projectId (required).
get_page_summaryA
Get a readable summary of a FlutterFlow page from local cache — widget tree, actions, params, state. Component references are resolved to show [ComponentName] (ComponentId) instead of plain Container. Use the ComponentId with get_component_summary to drill into a component. No API calls. Run sync_project first if not cached.
Inputs: projectId (required), pageName, scaffoldId.
get_project_setupA
Get Project Setup settings — Firebase services, Languages, Platforms, Permissions, Project Dependencies, Dev Environments. Mirrors the FlutterFlow 'Project Setup' settings section. Cache-based, no API calls. Run sync_project first.
Inputs: projectId (required).
get_project_yamlA
Read YAML files from the local project cache. Requires sync_project to be run first. Returns one file if fileName is specified, or lists all cached file keys if omitted.
Inputs: projectId (required), fileName.
get_themeA
Get theme colors, typography, breakpoints, and widget defaults from local cache. No API calls. Run sync_project first if not cached.
Inputs: projectId (required), section.
get_yaml_docsA
Search and retrieve FlutterFlow YAML reference documentation. Use topic to search by keyword (e.g. 'Button', 'actions', 'theming') or file to fetch a specific doc file. Returns the full doc content.
Inputs: file, topic.
list_pagesA
List all pages in a FlutterFlow project with human-readable names, scaffold IDs, and folder assignments. Use this FIRST to discover pages before fetching their YAML.
Inputs: projectId (required).
list_project_filesA
List all YAML file names in a FlutterFlow project. Supports optional prefix filter (e.g. 'page/', 'component/') to narrow results.
Inputs: projectId (required), prefix.
list_projectsA
List FlutterFlow projects for the authenticated user. NOTE: This may not return all projects you have access to (shared/team projects can be missing). If a project is missing, copy its ID directly from the FlutterFlow editor (click the project name in the top-left corner).
Inputs: project_type.
search_project_filesA
Search cached project file keys by keyword, prefix, or regex. Returns matching file keys for use with get_project_yaml. Cache-only, no API calls. Run sync_project first.
Inputs: query (required), projectId (required), mode.
sync_projectA
Sync an entire FlutterFlow project to the local cache. Downloads all YAML files (bulk or batched fallback) for fast offline reads. Use force=true to re-sync an already cached project.
Inputs: projectId (required), force.
update_project_yamlA
Push YAML changes to a FlutterFlow project. IMPORTANT: Always call validate_yaml first to check for errors before updating. For best results, call get_editing_guide before writing YAML to get the correct workflow and schema documentation.
Inputs: projectId (required), fileKeyToContent (required).
validate_yamlA
Validate YAML content before pushing changes to a FlutterFlow project. Always call this before update_project_yaml. Tip: Call get_editing_guide or get_yaml_docs BEFORE writing YAML to understand the correct schema and field names. Validation catches syntax errors but not semantic mistakes.
Inputs: fileKey (required), projectId (required), fileContent (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 25 tools
Tools are mostly distinct, with clear separation between listing, reading, searching, and updating. Some overlap exists among the many settings getters and between get_editing_guide and get_yaml_docs, but descriptions are specific enough to avoid major confusion.
All tools use consistent snake_case verb_noun naming with uniform prefixes like list_, get_, find_, validate_, update_, sync_, and search_. The pattern is predictable and makes the toolset easy to navigate.
At 25 tools, this is on the heavy side and right at the borderline for what feels reasonable. Each tool covers a specific FlutterFlow feature or settings section, so the count is justifiable, but it is still at the upper edge of appropriateness.
The tool set covers the full read/update workflow for YAML, including validation, documentation, and reference. Missing explicit tools like list_components or project creation/deletion, but these can be worked around via file listing and YAML updates, so gaps are minor.
Maintenance
Related MCP Connectors
Read and write your Fresh Jots notes from Claude, Cursor, and any MCP client.
Pageform is an AI-native data room platform with a remote MCP server for Claude and other MCP-compatible clients. Upload and organize documents, generate structured data rooms with AI, audit missing materials, create secure share links, and analyze investor engagement directly from your AI assistant. Built for fundraising, investor relations, diligence, and deal workflows. Connect securely through Pageform OAuth. Website: https://pageform.io Documentation: https://pageform.io/docs/integrations/claude
Share one project context across ChatGPT, Claude, Telegram and any MCP client.
Official Fibery MCP server: query and update your workspace from any AI client via OAuth.
Related MCP Servers
- AlicenseCqualityBmaintenanceA Model Context Protocol server that connects Flutter apps with AI coding assistants like Cursor, Claude, and Cline, enabling AI-powered analysis of widget trees, navigation, and layout issues.62399MIT
- AlicenseNot gradedqualityDmaintenanceIntegration project for Model Context Protocol (MCP) servers with Claude Desktop App, enabling filesystem operations, development support, and file management through natural language.MIT
- FlicenseNot gradedqualityDmaintenanceProvides filesystem access to Claude via the MCP protocol, enabling local file operations through natural language.-
- AlicenseNot gradedqualityDmaintenanceRead-only Figma MCP server that enables design-to-code workflows by talking to the Figma REST API with a personal access token, for use with Claude Code and GitHub Copilot.2,146 npmMIT