MCP Server

flet-pkg includes a Model Context Protocol (MCP) server that exposes scaffolding and code generation capabilities to AI agents like Claude.

Installation

The MCP server is included with flet-pkg. You also need the mcp dependency:

uv add flet-pkg[mcp]

Running standalone

flet-pkg-mcp

This starts the server using stdio transport (the standard for MCP).

Configuration

Claude Code

Add to your project's .mcp.json or ~/.claude/settings.json:

{
  "mcpServers": {
    "flet-pkg": {
      "command": "uv",
      "args": ["run", "--from", "flet-pkg[mcp]", "flet-pkg-mcp"]
    }
  }
}

Claude Desktop

Add to your Claude Desktop configuration file:

macOSWindows

~/Library/Application Support/Claude/claude_desktop_config.json

%APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "flet-pkg": {
      "command": "uv",
      "args": ["run", "--from", "flet-pkg[mcp]", "flet-pkg-mcp"]
    }
  }
}

Tools

The server exposes 7 tools:

tool_derive_names

Derive project/package/control names from a Flutter package name.

Parameters:

Name	Type	Description
flutter_package	str	Flutter package name (e.g. onesignal_flutter)

Returns: project_name, package_name, control_name, control_name_snake

tool_map_dart_type

Convert a Dart type string to its Python/Flet equivalent.

Parameters:

Name	Type	Default	Description
dart_type	str		Dart type to convert
flet_aware	bool	False	Use native Flet types for UI controls

tool_fetch_metadata

Get pub.dev package metadata (version, description, homepage).

Parameters:

Name	Type	Description
flutter_package	str	Flutter package name

tool_detect_extension_type

Auto-detect whether a Flutter package should be a service or ui_control extension.

Parameters:

Name	Type	Description
flutter_package	str	Flutter package name

Returns: extension_type ("service" or "ui_control")

tool_scaffold

Create a project skeleton from a template. Does not generate code — use tool_run_pipeline after this.

Parameters:

Name	Type	Default	Description
template_name	str		service or ui_control
flutter_package	str		Flutter package name
project_name	str		Project directory name
package_name	str		Python package name
control_name	str		PascalCase control class name
description	str	""	Package description
author	str	""	Author name
output_dir	str	"."	Output directory path
include_console	bool	True	Include debug console module

tool_run_pipeline

Run the full generation pipeline: download → parse → analyze → generate → write.

Parameters:

Name	Type	Default	Description
flutter_package	str		Flutter package name
control_name	str		PascalCase control class name
extension_type	str		service or ui_control
project_dir	str		Path to scaffolded project
package_name	str		Python package name
description	str	""	Package description
control_name_snake	str	""	Snake-case control name
include_console	bool	True	Include debug console module
local_package_path	str \| None	None	Local Flutter package path

tool_analyze_gaps

Analyze coverage gaps between a Flutter package and its generated code. Purely deterministic — no LLM required.

Parameters:

Name	Type	Description
flutter_package	str	Flutter package name
extension_type	str	service or ui_control

Returns: coverage_pct, total_dart_api, total_generated, feasible_gaps, and a list of gaps.

Resources

flet-pkg://type-map

Full Dart-to-Python type mapping table, including standard mappings, Flet-aware mappings, and skipped types.

flet-pkg://templates

List of available template names and their paths.

Prompts

scaffold_service

Step-by-step guide to scaffold a Flet service extension. Takes flutter_package as input.

scaffold_ui_control

Step-by-step guide to scaffold a Flet UI control extension. Takes flutter_package as input.

analyze_package

Guide to analyze a Flutter package's coverage without scaffolding. Takes flutter_package and extension_type.

Example prompts

With the MCP server configured, you can use natural language prompts in Claude Code or Claude Desktop:

Analyze a service package

> Use the flet-pkg tool to analyze the Flutter package "onesignal_flutter" and tell me
  the extension type and coverage.

Claude will call:

1. tool_detect_extension_type("onesignal_flutter") → service
2. tool_analyze_gaps("onesignal_flutter", "service") → Coverage 95.4%

Analyze a UI control package

> Use the flet-pkg tool to analyze the Flutter package "rive" and tell me the extension
  type and coverage.

Claude will call:

1. tool_detect_extension_type("rive") → ui_control
2. tool_analyze_gaps("rive", "ui_control") → Coverage report

Scaffold a full project

> Use flet-pkg to scaffold a service extension for shared_preferences.

Claude will call:

1. tool_derive_names("shared_preferences")
2. tool_fetch_metadata("shared_preferences")
3. tool_scaffold(template_name="service", ...)
4. tool_run_pipeline(extension_type="service", ...)