Code Writing Team

An automated AI-powered software development system that processes Jira tickets and implements code changes using a team of specialized AI agents.

Overview

This project automates the entire software development lifecycle from ticket intake to PR creation using a team of AI agents powered by Model Context Protocol (MCP) and Anthropic's Claude models. The system can:

1. Read Jira tickets and extract requirements
2. Clone relevant repositories
3. Organize tasks and create implementation plans
4. Write code according to specifications
5. Review code for quality and correctness
6. Create pull requests with all required changes
7. Update documentation in Confluence

Architecture

The system operates with specialized AI agents:

• Task Organizer: Analyzes tickets, creates task plans, and manages workflow
• Coder: Implements code changes based on requirements
• Code Reviewer: Reviews code for quality, correctness and best practices

The workflow is orchestrated by the MCP Orchestrator, which manages agent interactions and task execution.

Prerequisites

• Python 3.10+
• uv package manager
• Jira and Confluence access
• GitHub access with appropriate permissions
• Anthropic API key (Claude 3.7 Sonnet model)

Installation

1. Clone this repository

git clone <repository-url>
cd code_writing_team

2. Install dependencies using uv

uv pip sync uv.lock

3. Set up configuration

Copy and configure your secrets file:

cp mcp_agent.secrets.example.yaml mcp_agent.secrets.yaml

Edit mcp_agent.secrets.yaml with your actual API keys and credentials.

Known Issues

⚠️ Important: The git-mcp server is currently non-functional. You will need to manually clone repositories into the repos/ folder before running the system against a ticket. Once cloned, the AI agents will detect the existing repositories and work with them without attempting to clone.

Example of manually cloning a repository:

# Create the repos directory if it doesn't exist
mkdir -p repos

# Clone the repository needed for your ticket
git clone https://github.com/username/repository.git repos/repository

Ensure the repository name matches what is specified in the Jira ticket for the AI agents to locate it correctly.

Usage

Run the system by providing a Jira ticket key:

python main.py PROJ-123

The system will:

1. Validate MCP tools
2. Process the ticket
3. Clone the relevant repository
4. Create a task file in the tasks/ directory
5. Implement code changes
6. Create a pull request
7. Update the task file with progress

Project Structure

• main.py: Entry point for the application
• agents/: Specialized AI agent implementations
  • code_reviewer.py: Code review agent
  • coder.py: Implementation agent
  • task_organizer.py: Task organization agent
• workflows/: Workflow orchestration modules
  • process_ticket.py: Main ticket processing workflow
  • clone_repository.py: Repository cloning workflow
  • ensure_task_file_exists.py: Task file management
• utils/: Utility modules
  • config.py: Configuration management
  • app_instance.py: Application instance management
• repos/: Cloned repositories for ticket implementation
• tasks/: Task files for each ticket

Configuration

The system uses three configuration files:

• mcp_agent.config.yaml: General configuration for MCP servers and settings
• mcp_agent.secrets.yaml: Sensitive credentials and API keys (not committed to version control)
• .env: Environment variables for Docker and local development

Environment Variables (.env)

The .env file contains configuration used for both local development and Docker:

# Default ticket ID (can be overridden when running docker-compose)
TICKET_ID=AI-8

# Git user configuration
GIT_AUTHOR_NAME=YourName
GIT_AUTHOR_EMAIL=your.email@example.com

Key variables:

• TICKET_ID: Default Jira ticket to process when running in Docker
• GIT_AUTHOR_NAME and GIT_AUTHOR_EMAIL: Git user identity for commits

When using Docker, you can override the ticket ID:

TICKET_ID=PROJ-123 docker-compose up

Development

Adding a new MCP server

1. Add the server configuration to mcp_agent.config.yaml
2. Add required environment variables to mcp_agent.secrets.yaml
3. Update the validate_mcp_tools.py to check for the new server

Adding a new agent

1. Create a new agent file in the agents/ directory
2. Implement the agent with specialized capabilities
3. Add the agent to the orchestrator in process_ticket.py

Troubleshooting

Common Issues

• Missing MCP tools: Ensure all required MCP servers are configured correctly
• API rate limits: Check for rate limiting in Jira, GitHub, or Anthropic APIs
• Permission issues: Verify API tokens have appropriate permissions

Logs

Logs are output to the console by default. Set the log level in mcp_agent.config.yaml.

License

[Add your license information here]

Contributing

[Add your contribution guidelines here]