MCP Agile Flow

A comprehensive system for managing AI-assisted agile development workflows with a modern, resource-based API using FastMCP.

Status

✅ Migration Complete: The migration from legacy server to FastMCP implementation is fully complete. All legacy code and tests have been removed.

Overview

The MCP Agile Flow project uses a resource-based approach with FastMCP from the official MCP SDK, focusing on:

• RESTful API Design - Clean, intuitive resource URIs for data access
• Resource-First Architecture - Optimized for data retrieval and state representation
• Action-Oriented Tools - Tools used only for operations that modify state

Key Features

• Agile Documentation: Generate and maintain comprehensive AI documentation
• Project Structure: Organize your project with AI-generated files and directories
• IDE Integration: Direct integration with various AI IDEs (Cursor, Windsurf, Cline)
• Workflow Management: Track agile stories and progress
• Intuitive API Structure: Resources organized in a RESTful hierarchy
• Simplified Integration: Direct mapping to resource URIs
• Improved Performance: Optimized for data access patterns

Getting Started

To use MCP Agile Flow:

1. Install the package:

   pip install mcp-agile-flow
   

2. Import in your code:

   from mcp_agile_flow import call_tool, call_tool_sync
   
   # Use async interface
   result = await call_tool("get-project-settings", {})
   
   # Or use sync interface
   result = call_tool_sync("get-project-settings", {})
   

MCP Client Configuration

Important: Configuration Update Required

If you had previously configured MCP Agile Flow, you need to update your configuration. The fastmcp_server.py module has been removed as part of code cleanup, and functionality has been consolidated into the main package.

Update your MCP client configuration from:

{
    "name": "mcp-agile-flow",
    "server": {
        "type": "module",
        "module": "mcp_agile_flow.fastmcp_server",
        "entry_point": "run"
    }
}

To:

{
    "name": "mcp-agile-flow",
    "server": {
        "type": "module",
        "module": "mcp_agile_flow",
        "entry_point": "main"
    }
}

For Cursor users, also update the mcp.json file (typically at ~/.cursor/mcp.json):

"mcp-agile-flow": {
  "command": "/path/to/python",
  "args": [
    "-m",
    "mcp_agile_flow"  // Updated from "mcp_agile_flow.fastmcp_server"
  ],
  "autoApprove": [
    // ...
  ]
}

Command Line Usage

You can also run the server directly from the command line:

# Using Python (logs disabled by default)
python -m mcp_agile_flow

# Enable normal logging
python -m mcp_agile_flow --verbose

# Debug mode (most verbose logging)
python -m mcp_agile_flow --debug

Available Tools

The MCP Agile Flow provides several tools:

• get-project-settings: Get project settings including paths and environment variables
• initialize-ide: Initialize project directory structure for specific IDEs
• initialize-ide-rules: Initialize AI rule files for specific IDEs
• prime-context: Analyze project documentation and build contextual understanding
• migrate-mcp-config: Migrate MCP configuration between different IDEs
• think: Record a thought for complex reasoning and step-by-step analysis
• get-thoughts: Retrieve all thoughts recorded in the current session
• clear-thoughts: Clear all recorded thoughts from the current session
• get-thought-stats: Get statistics about the thoughts recorded in the current session
• process-natural-language: Process natural language commands and route to appropriate tools

Natural Language Commands

MCP Agile Flow supports natural language commands, making it easier to interact with the tools without remembering exact command names. Simply use conversational phrases, and the system will automatically detect your intent and map them to the appropriate tools with the correct parameters.

Supported Command Types

Migration Commands

To migrate MCP configuration between different IDEs:

• "migrate mcp config to claude-desktop"
• "migrate config from cursor to claude-desktop"
• "copy mcp settings to windsurf"
• "transfer config to cline"
• "move mcp settings from cursor to roo"

If the source IDE is not specified, it defaults to "cursor".

Note: Valid IDE names are: "cursor", "windsurf-next", "windsurf", "cline", "roo", and "claude-desktop".

Initialization Commands

To initialize a project with rules for a specific IDE:

• "initialize ide for claude"
• "setup rules for windsurf"
• "create ide for cline"
• "initialize rules for copilot"

Project Settings Commands

To get comprehensive project settings:

• "get project settings"
• "show settings"
• "project settings"

Context Analysis Commands

To analyze project documentation:

• "prime context"
• "analyze project context"
• "build context"

Thinking Commands

To record a thought:

• "think about [your thought here]"

Usage Examples

Here are some examples of how to use these commands:

from mcp_agile_flow import process_natural_language

# Migrate configuration from Cursor to Claude
result = process_natural_language("migrate mcp config to claude-desktop")

# Initialize rules for Windsurf
result = process_natural_language("initialize ide for windsurf")

# Get project settings
result = process_natural_language("get project settings")

# Prime the context
result = process_natural_language("prime context")

# Record a thought
result = process_natural_language("think about how to improve code quality")

Using from Command Line

You can also use natural language commands with the MCP Agile Flow CLI:

python -m mcp_agile_flow process-natural-language "migrate mcp config to claude-desktop"

Error Handling

If the system cannot recognize a command, it will return an error message explaining that no command was detected and suggesting to use more specific wording.

Extending Commands

The natural language command detection is implemented in utils.py using regular expressions. To add support for new command patterns, add appropriate regex patterns to the detect_mcp_command function.

Development

To set up for development:

1. Clone the repository:

   git clone https://github.com/yourusername/mcp-agile-flow.git
   cd mcp-agile-flow
   

2. Create a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

3. Install development dependencies:

   pip install -e ".[dev]"
   

4. Run tests:

   pytest
   

5. Common Makefile commands:

   make test             # Run all tests
   make test-nl-commands # Test natural language command functionality
   make test-core        # Run core tests only
   make coverage         # Generate coverage report
   make clean            # Clean build artifacts
   make clean-all        # Clean everything including venv
   make clean-archived   # Remove archived legacy files
   

License

This project is licensed under the MIT License - See LICENSE file for details.