ClaudeAutoPM

Automated project management system to ship faster better using spec-driven development and AI agents running in parallel

Stop losing context. Stop blocking on tasks. Stop shipping bugs. ClaudeAutoPM transforms PRDs into epics, epics into issues, and issues into production code – with full traceability at every step.

📖 Documentation

Full documentation is available at: https://rafeekpro.github.io/ClaudeAutoPM/

• 📚 Complete Guide
• 📝 Command Reference
• 🤖 Agent Registry
• 🔧 Development Guide
• ⚙️ Configuration

🎯 Key Features

Hybrid Command System

ClaudeAutoPM uses a hybrid approach combining deterministic operations with AI-powered intelligence:

Core Capabilities

• 96+ Professional CLI Commands - Full yargs-based CLI with comprehensive PM commands
• Hybrid Execution - Choose between templates (deterministic) or AI assistance
• AI Agent Integration - Parallel execution with specialized agents (in Claude Code)
• Dynamic Agent Teams - Switch between specialized agent teams based on context (DevOps, Frontend, Backend, Fullstack)
• Azure DevOps & GitHub - Complete project management integration
• Smart Context Management - Never lose track of your work
• Automated Workflows - From PRD to production deployment

📺 Visual Walkthrough

See ClaudeAutoPM in action - from installation to deployment. Click to expand each video:

1️⃣ Install AutoPM - Complete installation process

2️⃣ First Claude Execution - Setting up and running Claude Code

3️⃣ Creation of PRD - Product Requirements Document workflow

4️⃣ GitHub Sync and Start Working - Synchronizing with GitHub and beginning tasks

5️⃣ Issues Finished - Completing and closing tasks

6️⃣ Checking the Work Done - Running Web App + FastAPI demo

🚀 Get Started in 5 Minutes

🎯 Quick Start for New Users

npm install -g claude-autopm

# Run the interactive guide
autopm guide

The interactive guide will walk you through:

• ✅ System requirements verification
• 📦 Installation with preset selection
• ⚙️ Provider configuration (GitHub/Azure DevOps/Local)
• 🤖 Agent team management and automatic switching
• 📝 Creating your first PRD and project workflow
• 🆘 Troubleshooting and diagnostics
• 📚 Complete documentation and resources

Manual Setup

1. Install (30 seconds)

npm install -g claude-autopm

2. Setup Your Project (2 minutes)

# Navigate to your project
cd your-project/

# Install ClaudeAutoPM framework
autopm install

During installation, you'll be asked to:

1. Choose a configuration preset:

💡 Note: The installer automatically detects available tools (Docker, kubectl) and only shows compatible options. Missing tools? See installation links in the interactive prompt.

2. Select your project management provider:

3. Update Existing Installation

# Update to latest framework version
autopm update

# Or force update with options
autopm update --force --no-backup

Update features:

• 🔄 Smart Updates - Preserves your configuration and project data
• 📦 Automatic Backup - Creates backup before updating (can be disabled)
• ⚙️ Config Preservation - Keeps your settings, teams, and provider config
• 📁 Data Protection - Preserves epics, PRDs, and all project files
• 🔧 Version Detection - Only updates when necessary

Update options:

• --force - Force update even if versions match
• --no-backup - Skip backup creation
• --no-preserve-config - Don't preserve configuration files

4. Initialize PM System (1 minute)

# In Claude Code, run:
/pm:init

# Create your CLAUDE.md with repository info:
/init include rules from .claude/CLAUDE.md

4.5 Provider Configuration (NEW)

# View current configuration
autopm config show

# Switch to Azure DevOps
autopm config set provider azure
autopm config set azure.organization your-org
autopm config set azure.project your-project
export AZURE_DEVOPS_PAT=your-token

# Switch to GitHub
autopm config set provider github
autopm config set github.owner your-username
autopm config set github.repo your-repo
export GITHUB_TOKEN=your-token

# Quick switch between providers
autopm config switch azure
autopm config switch github

# Validate configuration
autopm config validate

4.6 Verify Installation & Configuration

After installation, verify that everything is properly configured:

# Comprehensive configuration check
autopm validate

# Shows:
# ✅ Essential Components (.claude directory, config, provider, git)
# ✅ Optional Components (MCP servers, git hooks, Node.js version)
# 📋 Next steps for incomplete setup

Example Output:

╔════════════════════════════════════════════════════════════════╗
║         🔍 ClaudeAutoPM Configuration Status                  ║
╚════════════════════════════════════════════════════════════════╝

Essential Components:
  ✅ .claude directory - Framework installed
  ✅ Configuration file - Provider: github
  ✅ Provider setup - GitHub (configured)
  ✅ Git repository - Initialized

Optional Components:
  ✅ MCP servers - 2 active (context7-docs, github-mcp)
  ⚠️  Git hooks - Not installed (run: bash scripts/setup-hooks.sh)
  ✅ Node.js version - v20.10.0 (compatible)

Next Steps:
  1. Install git hooks: bash scripts/setup-hooks.sh
  2. Run MCP configuration check: autopm mcp check

4. Ship Your First Feature (90 seconds)

Option A: Using Templates (Works Everywhere)

# Create PRD from template
autopm pm:prd-new user-auth --template

# Convert PRD to basic epic structure
autopm pm:prd-parse user-auth --basic

# Generate tasks from template
autopm pm:epic-decompose user-auth --template backend

# Push to GitHub/Azure DevOps
autopm pm:epic-sync user-auth

Option B: AI-Powered in Claude Code

# Create PRD through AI brainstorming
/pm:prd-new user-auth

# AI-powered PRD analysis and epic creation
/pm:prd-parse user-auth

# Intelligent task decomposition
/pm:epic-decompose user-auth

# Push to GitHub and start building
/pm:epic-sync user-auth

That's it! You're now using structured, spec-driven development with AI agents.

🎯 Core Workflow: Crawl → Walk → Run

Crawl: Basic Commands

/pm:prd-new feature        # Create product requirements
/pm:epic-oneshot feature   # Plan, decompose, and sync to GitHub
/pm:issue-start 1234       # Start working on a task
/pm:next                   # Get next priority task

Walk: Team Collaboration

/pm:status                 # Project dashboard
/pm:standup               # Daily standup report
/pm:issue-sync 1234       # Push progress to GitHub
/pm:blocked               # Show blocked tasks

Run: Advanced Features

• 50+ Specialized AI Agents - Python, React, K8s, AWS, and more
• Parallel Execution - Multiple agents working simultaneously
• Cross-Platform - GitHub, Azure DevOps, GitLab (coming soon)
• MCP Integration - Model Context Protocol for live documentation, codebase analysis, and tool integration
• Agent Analysis - Discover which agents use MCP and their configurations
• Interactive Diagnostics - Comprehensive health checks and connection testing

🤖 Dynamic Agent Teams

ClaudeAutoPM now supports dynamic agent teams that can be switched based on your current context:

# List available teams
autopm team list

# Load a specialized team
autopm team load devops        # CI/CD and infrastructure agents
autopm team load frontend      # React, JavaScript, UX agents
autopm team load python_backend # Python, FastAPI, Flask agents
autopm team load fullstack     # Combined frontend + backend

# Check current team
autopm team current

🚀 Automatic Team Switching

ClaudeAutoPM can automatically switch teams based on your Git branch name!

Setup (one-time):

# Enable automatic team switching
bash scripts/setup-githooks.sh

Branch Naming Convention:

# Branch pattern: type/team/description
git checkout -b feature/devops/add-ci      # Auto-loads 'devops' team
git checkout -b fix/frontend/button-style  # Auto-loads 'frontend' team
git checkout -b feat/backend/new-api       # Auto-loads 'python_backend' team

Available Teams:

• base - Core agents for general tasks
• devops - Docker, Kubernetes, GitHub Operations, Terraform
• frontend - React, JavaScript, E2E testing, UX design
• python_backend - FastAPI, Flask, PostgreSQL, MongoDB
• fullstack - Inherits from both frontend and python_backend

Teams support inheritance, so specialized teams automatically include base agents. The active team configuration is saved in your project and persists across sessions.

🔌 MCP (Model Context Protocol) Management

ClaudeAutoPM provides comprehensive MCP management for agent-to-tool integration. 39 out of 53 agents (74%) use MCP servers for live documentation and external tool access.

Quick Start with MCP

# List all available MCP servers
autopm mcp list

# Check which agents use MCP
autopm mcp agents
🤖 Agents Using MCP

✅ react-frontend-engineer
   └─ context7-docs

✅ python-backend-engineer
   └─ context7-docs
   └─ sqlite-mcp

📊 Summary:
   Total agents: 53
   Using MCP: 39 (74%)

# Enable MCP servers
autopm mcp enable context7-docs
autopm mcp enable github-mcp

# Configure API keys interactively
autopm mcp setup

# Sync configuration
autopm mcp sync

MCP Commands

Agent Analysis

autopm mcp agents              # List agents using MCP
autopm mcp agents --by-server  # Group by MCP server
autopm mcp agent <name>        # Show agent MCP config
autopm mcp usage               # Usage statistics
autopm mcp tree                # Dependency tree

Configuration & Diagnostics

autopm mcp check               # Quick configuration check
autopm mcp setup               # Interactive API key setup
autopm mcp diagnose            # Run full diagnostics
autopm mcp test <server>       # Test server connection
autopm mcp status              # Show servers status

Server Management

autopm mcp list                # List available servers
autopm mcp info <server>       # Server details
autopm mcp enable <server>     # Enable server
autopm mcp disable <server>    # Disable server
autopm mcp sync                # Sync configuration

MCP Server Types

• Documentation - context7-docs for live framework documentation (React, Python, AWS, etc.)
• Codebase - context7-codebase for project analysis and navigation
• GitHub - github-mcp for repository operations
• Databases - sqlite-mcp, postgresql-mcp, mongodb-mcp for data operations
• Browser - playwright-mcp for E2E testing and automation

Example: Setting Up Context7

# 1. Enable context7 documentation server
autopm mcp enable context7-docs

# 2. Configure API key
cat > .claude/.env << EOF
CONTEXT7_API_KEY=your-api-key-here
CONTEXT7_WORKSPACE=your-workspace-id
EOF

# 3. Sync configuration
autopm mcp sync

# 4. Verify setup
autopm mcp diagnose

# 5. Test connection
autopm mcp test context7-docs

MCP Benefits

• ✅ Live Documentation - Agents access latest framework docs automatically
• ✅ Codebase Context - Deep understanding of your project structure
• ✅ Tool Integration - GitHub, databases, browsers, and more
• ✅ No Hallucinations - Real-time data instead of outdated training
• ✅ Extensible - Add custom MCP servers for your tools

📖 Full MCP Setup Guide: MCP_SETUP_GUIDE.md

📚 Full Documentation

For comprehensive guides, advanced features, and detailed configuration:

👉 Visit the ClaudeAutoPM Wiki

CLI System (96 Commands)

ClaudeAutoPM features a professional CLI powered by yargs with 96 commands:

# View all commands
autopm --help

# Command categories:
# - Azure DevOps (39 commands)
# - Project Management (32 commands)
# - AI & Automation (2 commands)
# - Infrastructure & DevOps (7 commands)
# - Testing & Context (5 commands)
# - UI & Frontend (6 commands)
# - Utilities (5 commands)

# Examples:
autopm pm:init --type web
autopm azure:sprint-status
autopm context:create

Latest Features (v1.12.x)

v1.12.2 - Smart Installation & Command Fixes

• Smart Tool Detection - Installer automatically detects Docker and kubectl availability
• Intelligent Defaults - Installation options filtered based on available tools
• Command Format Fixes - All PM scripts now use consistent /pm:... format
• Version Tracking - Update command now properly detects installed versions

v1.9.x - Complete PM Command Suite

17 new PM commands providing comprehensive project management capabilities:

PRD & Epic Management

# Create new Product Requirements Document
autopm pm:prd-new user-authentication --description "OAuth 2.0 integration"

# Parse PRD into executable epic with tasks
autopm pm:prd-parse user-authentication --overwrite

# Close completed epic
autopm pm:epic-close user-auth "All authentication features complete"

Splitting Large PRDs into Multiple Epics

For complex features, you can split a single PRD into multiple focused epics:

# 1. Create a comprehensive PRD
autopm pm:prd-new payment-system --template

# 2. Split into focused epics using AI analysis
/pm:prd-split payment-system

# This creates multiple epics from one PRD:
# ├─ payment-system-backend     (API, database, payment gateway integration)
# ├─ payment-system-frontend    (UI components, checkout flow)
# └─ payment-system-security    (PCI compliance, encryption, audit logging)

# 3. Work on each epic independently
/pm:epic-decompose payment-system-backend
/pm:epic-sync payment-system-backend

# 4. Track overall PRD progress
autopm pm:prd-status payment-system
# Shows:
#   Epic: payment-system-backend  [████████░░] 80% (8/10 tasks)
#   Epic: payment-system-frontend [████░░░░░░] 40% (4/10 tasks)
#   Epic: payment-system-security [░░░░░░░░░░]  0% (0/8 tasks)
#   Overall: 44% complete (12/28 tasks)

When to split PRDs:

• ✅ Feature requires multiple specialized teams (frontend, backend, DevOps)
• ✅ Different components have separate deployment timelines
• ✅ Epic would exceed 15-20 tasks (becomes hard to manage)
• ✅ Clear architectural boundaries exist (UI, API, infrastructure)

Best practices:

• 📝 Keep original PRD as the source of truth
• 🏷️ Use consistent naming: <prd-name>-<component>
• 🔗 Link epics back to parent PRD in description
• 📊 Track overall progress across all child epics
• 🎯 Each epic should be independently deployable when possible

Issue Lifecycle

# Start working on an issue
autopm pm:issue-start ISSUE-123

# Show issue details and progress
autopm pm:issue-show ISSUE-123

# Edit issue details
autopm pm:issue-edit ISSUE-123 --title "Updated: Fix authentication bug"

# Close completed issue
autopm pm:issue-close ISSUE-123 "Fixed OAuth redirect issue"

Pull Request Workflow

# Create PR with auto-generated description
autopm pm:pr-create "feat: implement OAuth authentication"

# List all open PRs
autopm pm:pr-list --status open

# Create draft PR for work in progress
autopm pm:pr-create "wip: user dashboard" --draft

Context Management

# Create development context for feature
autopm pm:context-create user-authentication

# Update context with current progress
autopm pm:context-update user-authentication

# Prime context for AI assistance
autopm pm:context-prime user-authentication

Project Maintenance

# Optimize project structure and reduce context size
autopm pm:optimize --apply

# Clean archived work and free up space
autopm pm:clean --archive-old

# Sync work state across team
autopm pm:sync --push-remote

# Create versioned release
autopm pm:release minor --tag "v1.2.0"

📖 Full CLI Documentation

Popular Topics

• Configuration Options
• Agent Registry (50+ agents)
• Command Reference
• Docker & Kubernetes
• Azure DevOps Integration

💡 What Makes This Different?

🔑 Key Principles

1. No Vibe Coding - Every line traces back to a specification
2. GitHub as Database - Issues are the single source of truth
3. Parallel by Default - Multiple agents, maximum velocity
4. Context Preservation - Never lose project state again

🛠️ System Requirements

• Node.js >= 16.0.0
• npm >= 8.0.0
• Git (system installation)
• Claude Code or compatible AI assistant

🧪 Testing

Running Tests

# Run all unit tests (default)
npm test

# Run Azure DevOps integration tests (requires Azure environment)
npm run test:azure:integration

Test Separation

Tests are separated into unit and integration categories:

• Unit tests - Run without external dependencies (default)
• Integration tests - Require Azure DevOps environment variables:
  • AZURE_DEVOPS_PAT - Personal Access Token
  • AZURE_DEVOPS_ORG - Organization name
  • AZURE_DEVOPS_PROJECT - Project name

To run integration tests, set AZURE_DEVOPS_INTEGRATION_TESTS=true along with the required environment variables.

📊 Configuration Preset Details

Feature Comparison

Execution Strategies

• Sequential: Safe, one agent at a time, predictable
• Adaptive: Intelligent parallelization based on task complexity (learning mode planned)
• Hybrid: Maximum parallelization, fastest but requires monitoring

When to Choose Each Preset

Minimal

• New to ClaudeAutoPM
• Small personal projects
• No Docker environment available
• Learning the system

Docker-only

• Medium-sized projects
• Local development focus
• Docker available but not Kubernetes
• Team of 2-10 developers

Full DevOps (Recommended)

• Production projects
• CI/CD pipelines required
• Enterprise environments
• Teams with DevOps practices
• Best balance of features and safety

Performance

• Large codebases (>100k LOC)
• Powerful development machines (16+ cores)
• Experienced teams
• When build speed is critical

What Gets Installed

All presets install the same core files:

• .claude/agents/ - AI agent definitions
• .claude/commands/ - PM command library
• .claude/rules/ - Development rules
• .claude/scripts/ - Helper scripts
• .claude/checklists/ - Development checklists
• .claude/mcp/ - Model Context Protocol configs

Optional components (based on preset):

• Git hooks → .git/hooks/ (Docker-only, Full DevOps, Performance)
• Docker configurations → .claude/docker/ (all except Minimal)
• Kubernetes manifests → .claude/k8s/ (Full DevOps, Performance)
• CI/CD templates → .github/workflows/ or .azure-pipelines/

Git Hooks Installation

Git hooks install in two phases:

1. Automatic enablement (based on preset):

   • Minimal: No hooks
   • Docker-only, Full DevOps, Performance: Hooks enabled

2. User confirmation (if enabled):

   Would you like to install git hooks for automated commit/push validation? (y/n)

   • Yes → Installs pre-commit and pre-push hooks
   • No → Skips, can install later with .claude/scripts/setup-hooks.sh

Provider Integration Details

GitHub Issues

• Creates issues and PRs via GitHub API
• Syncs epics as GitHub milestones
• Requires: GITHUB_TOKEN with repo scope
• Commands: All pm:* plus github:* commands

Azure DevOps ✨ NEW: Full 3-Level Hierarchy Support

• Epic → User Story → Task hierarchy (full Azure DevOps model)
• Integrates with Azure Boards work items
• Automatic parent-child linking of work items
• Syncs with Azure Pipelines
• Requires: AZURE_DEVOPS_PAT with Work Items scope
• Commands: All pm:* plus azure:* commands

Azure DevOps Workflow Example:

# Configure for Azure DevOps
autopm config set provider azure
autopm config set azure.organization mycompany
autopm config set azure.project myproject

# Create and decompose PRD (creates Epic → User Stories → Tasks)
autopm pm:prd-new payment-gateway
autopm pm:epic-decompose payment-gateway  # Creates 3-level hierarchy

# Sync to Azure DevOps - creates linked work items
autopm pm:epic-sync payment-gateway
# ✅ Created Epic #1234 "Payment Gateway Integration"
# ✅ Created User Story #1235 "As a user, I want to pay with credit card"
#    ✅ Created Task #1236 "Implement Stripe API integration"
#    ✅ Created Task #1237 "Add payment form validation"
# ✅ Created User Story #1238 "As a user, I want to see payment history"
#    ✅ Created Task #1239 "Create payment history endpoint"
#    ✅ Created Task #1240 "Build payment history UI"

# Work items are automatically linked with parent-child relationships
# View in Azure Boards to see the full hierarchy visualization

Skip for now

• All data stored locally in .claude/
• No external synchronization
• Can migrate to provider later
• Perfect for evaluation or offline work

🌟 Support This Project

If ClaudeAutoPM helps your team ship better:

• ⭐ Star this repository
• 📦 Try on npm
• 🐦 Follow @rafeekpro on X

Built by developers who ship, for developers who ship.

Inspired by CCPM - Building upon the foundation of AI-powered project management.