Claude MPM - Multi-Agent Project Manager

A powerful orchestration framework for Claude Code (CLI) that enables multi-agent workflows, session management, and real-time monitoring through a streamlined Rich-based interface.

⚠️ Important: Claude MPM requires Claude Code CLI (v1.0.92+), not Claude Desktop (app). All MCP integrations work with Claude Code's CLI interface only.

Don't have Claude Code? Install from: https://docs.anthropic.com/en/docs/claude-code

Version Requirements:

Minimum: v1.0.92 (hooks support)

Recommended: v2.0.30+ (latest features)

Quick Start: See docs/user/getting-started.md to get running in 5 minutes!

🚀 What's New in v5.0

Git Repository Integration for Agents & Skills

Claude MPM now supports Git repositories for distributing and managing agents and skills - the most requested feature from our community!

🎯 Key Highlights:

📦 Massive Library: 47+ agents and hundreds of skills deployed automatically from curated repositories
🏢 Official Content: Anthropic's official skills repository included by default
🔧 Fully Extensible: Add your own agent/skill repositories with immediate testing
🌳 Smart Organization: Hierarchical BASE-AGENT.md inheritance for shared templates
📊 Clear Visibility: Two-phase progress bars show sync and deployment status
✅ Fail-Fast Testing: Test repositories before they cause startup issues

Quick Start: Add Your Own Repository

# Add custom agent repository (with immediate testing)
claude-mpm agent-source add https://github.com/yourorg/your-agents

# Add custom skill repository
claude-mpm skill-source add https://github.com/yourorg/your-skills

# Test repository without saving
claude-mpm agent-source add https://github.com/yourorg/your-agents --test

# List all configured sources
claude-mpm agent-source list

Default Repositories

Agents (47+ agents):

🏢 System: bobmatnyc/claude-mpm-agents - Core agent templates

Skills (hundreds of skills):

🏢 System: bobmatnyc/claude-mpm-skills - Community skills
🎯 Official: anthropics/skills - Anthropic's official skills (14+ skills)

What This Means:

✅ Automatic Setup: Everything deploys on first run
✅ Always Updated: Git repositories sync on startup
✅ Nested Support: Repositories with subdirectories are automatically flattened
✅ Template Inheritance: BASE-AGENT.md files cascade from parent directories
✅ Priority System: Control which sources win when multiple repos provide the same agent/skill

Hierarchical Organization Example

Support for BASE-AGENT.md enables powerful template inheritance:

your-agents/
  BASE-AGENT.md              # Root shared content (all agents inherit)
  engineering/
    BASE-AGENT.md            # Engineering-specific (engineering/* agents inherit)
    python/
      fastapi-engineer.md    # Inherits both BASE files

See docs/features/hierarchical-base-agents.md for complete guide.

Features

🎯 Multi-Agent Orchestration

47+ Specialized Agents: Comprehensive coverage from Git repositories (Python, Rust, QA, Security, Ops, etc.)
Smart Task Routing: PM agent intelligently delegates work to specialist agents
Session Management: Resume previous sessions with --resume
Resume Log System: Proactive context management with automatic 10k-token session logs at 70%/85%/95% thresholds

📦 Git Repository Integration (NEW in v5.0)

🏢 Curated Content: Default repositories with 47+ agents and hundreds of skills
🎯 Official Skills: Anthropic's official skills included by default
🔧 Custom Repositories: Add your own via CLI or configuration
🌳 Nested Support: Automatic flattening of nested directory structures
📝 Template Inheritance: BASE-AGENT.md for hierarchical organization
✅ Immediate Testing: Fail-fast validation when adding repositories
📊 Progress Visibility: Two-phase progress bars (sync + deployment)
⚡ Smart Caching: ETag-based caching reduces bandwidth by 95%+

🎯 Skills System

Hundreds of Skills: Comprehensive skill library from curated repositories
Official Skills: 14+ skills from Anthropic's official repository
Three-Tier Organization: Bundled/user/project skills with priority resolution
Auto-Linking: Skills automatically linked to relevant agents
Interactive Configuration: Skills wizard via claude-mpm configure
Version Tracking: Semantic versioning for all skills

🔌 Advanced Integration

MCP Integration: Full support for Model Context Protocol services
Real-Time Monitoring: Live dashboard with --monitor flag
Multi-Project Support: Per-session working directories
Git Integration: View diffs and track changes across projects

⚡ Performance & Security

Simplified Architecture: ~3,700 lines removed for better performance
Enhanced Security: Comprehensive input validation and sanitization
Intelligent Caching: File-based instruction caching for optimal performance
- ✅ Eliminates ARG_MAX limits on Linux (exceeds by 19.1%) and Windows (exceeds by 476%)
- ✅ ~200ms faster agent startup with hash-based cache invalidation
- ✅ Automatic and transparent (no configuration required)

Quick Installation

Prerequisites

Before installing Claude MPM, ensure you have:

Python 3.11+ (required for kuzu-memory dependency)
Claude Code CLI v1.0.92+ (required!)

# Verify Claude Code is installed
claude --version

# If not installed, get it from:
# https://docs.anthropic.com/en/docs/claude-code

Install Claude MPM

# Basic installation
pip install claude-mpm

# Or with uv (faster)
uv tool install claude-mpm

# Install with monitoring dashboard (recommended)
pip install "claude-mpm[monitor]"
uv tool install "claude-mpm[monitor]"

Or with pipx/uvx (recommended for isolated installation):

# Basic installation
pipx install claude-mpm

# Or with uv tool (permanent installation)
uv tool install claude-mpm

# Or with uvx (one-off execution)
uvx claude-mpm

# Install with monitoring dashboard (recommended)
pipx install "claude-mpm[monitor]"
uv tool install "claude-mpm[monitor]"

Verify Installation

# Check versions
claude-mpm --version
claude --version

# Run diagnostics (checks Claude Code compatibility)
claude-mpm doctor

# Verify Git repositories are synced
ls ~/.claude/agents/    # Should show 47+ agents
ls ~/.claude/skills/    # Should show hundreds of skills

# Check agent sources
claude-mpm agent-source list

# Check skill sources
claude-mpm skill-source list

What You Should See:

✅ 47+ agents deployed to ~/.claude/agents/
✅ Hundreds of skills deployed to ~/.claude/skills/
✅ Two agent sources configured (system + official)
✅ Two skill sources configured (system + official)
✅ Progress bars showing sync and deployment phases

💡 Optional Dependencies:

[monitor] - Full monitoring dashboard with Socket.IO and async web server components
[mcp] - Additional MCP services (mcp-browser, mcp-ticketer) - most users won't need this

🎉 Pipx Support Now Fully Functional! Recent improvements ensure complete compatibility:

✅ Socket.IO daemon script path resolution (fixed)
✅ Commands directory access (fixed)
✅ Resource files properly packaged for pipx environments
✅ Python 3.13+ fully supported

🤝 Recommended Partner Products

Claude MPM works excellently with these complementary MCP tools. While optional, we strongly recommend installing them for enhanced capabilities:

kuzu-memory - Advanced Memory Management

What it does: Provides persistent, project-specific knowledge graphs that enable agents to learn and retain context across sessions. Your agents will remember project patterns, architectural decisions, and important context automatically.

Installation:

pipx install kuzu-memory
# Or with uv
uv tool install kuzu-memory

Benefits with Claude MPM:

🧠 Persistent Context: Agents remember project-specific patterns and decisions across sessions
🎯 Intelligent Prompts: Automatically enriches agent prompts with relevant historical context
📊 Knowledge Graphs: Structured storage of project knowledge, not just flat memory
🔄 Seamless Integration: Works transparently in the background with zero configuration
💡 Smart Learning: Agents improve over time as they learn your project's patterns

Perfect for: Long-running projects, teams needing consistent context, complex codebases with deep architectural patterns.

Learn more: kuzu-memory on PyPI | GitHub Repository

mcp-vector-search - Semantic Code Search

What it does: Enables semantic code search across your entire codebase using AI embeddings. Find code by what it does, not just what it's named. Search for "authentication logic" and find relevant functions even if they're named differently.

Installation:

pipx install mcp-vector-search
# Or with uv
uv tool install mcp-vector-search

Benefits with Claude MPM:

🔍 Semantic Discovery: Find code by intent and functionality, not just keywords
🎯 Context-Aware: Understand code relationships and similarities automatically
⚡ Fast Indexing: Efficient vector embeddings for large codebases
🔄 Live Updates: Automatically tracks code changes and updates index
📊 Pattern Recognition: Discover similar code patterns and potential refactoring opportunities

Use with: /mpm-search "authentication logic" command in Claude Code sessions or claude-mpm search CLI command.

Perfect for: Large codebases, discovering existing functionality, finding similar implementations, architectural exploration.

Learn more: mcp-vector-search on PyPI | GitHub Repository

Quick Setup - Both Tools

Install both recommended tools in one go:

pipx install kuzu-memory
pipx install mcp-vector-search

# Or with uv
uv tool install kuzu-memory
uv tool install mcp-vector-search

Then verify they're working:

claude-mpm verify

That's it! These tools integrate automatically with Claude MPM once installed. No additional configuration needed.

That's it! See docs/user/getting-started.md for immediate usage.

Cache Management

Claude MPM maintains a local cache of agent templates at ~/.claude-mpm/cache/agents/.

Cache Structure

~/.claude-mpm/cache/agents/
└── bobmatnyc/
    └── claude-mpm-agents/
        ├── .git/              # Git repository
        ├── agents/            # 45+ agent templates
        └── docs/

Note: Prior to v5.4.23, agents were cached to ~/.claude-mpm/cache/remote-agents/. This has been standardized to ~/.claude-mpm/cache/agents/. Existing caches are automatically migrated.

Git Workflow (Optional)

If your cache is a git repository, you can manage agents with git operations:

# Check cache status
claude-mpm agents cache-status

# Pull latest agents
claude-mpm agents cache-pull

# Commit local changes
claude-mpm agents cache-commit --message "feat: update agents"

# Push changes
claude-mpm agents cache-push

# Full sync workflow
claude-mpm agents cache-sync

Or use Makefile targets:

make agents-cache-status   # Show git status
make agents-cache-pull     # Pull latest
make agents-cache-sync     # Full sync (pull + commit + push)
make deploy-agents         # Deploy with auto-pull

HTTP Sync Fallback

If cache is not a git repository, Claude MPM automatically falls back to HTTP sync with GitHub API.

Migration from Legacy Cache

If you have an old cache/agents/ directory, run the migration script:

python scripts/migrate_cache_to_remote_agents.py

See also: docs/CACHE_MANAGEMENT.md for comprehensive cache management guide.

Quick Usage

# Start interactive mode (recommended)
claude-mpm

# Start with monitoring dashboard
claude-mpm run --monitor

# Use semantic code search (auto-installs mcp-vector-search on first use)
claude-mpm search "authentication logic"
# or inside Claude Code session:
/mpm-search "authentication logic"

# Use MCP Gateway for external tool integration
claude-mpm mcp

# Run comprehensive health diagnostics
claude-mpm doctor

# Generate detailed diagnostic report with MCP service analysis
claude-mpm doctor --verbose --output-file doctor-report.md

# Run specific diagnostic checks including MCP services
claude-mpm doctor --checks installation configuration agents mcp

# Check MCP service status specifically
claude-mpm doctor --checks mcp --verbose

# Verify MCP services installation and configuration
claude-mpm verify

# Auto-fix MCP service issues
claude-mpm verify --fix

# Verify specific service
claude-mpm verify --service kuzu-memory

# Get JSON output for automation
claude-mpm verify --json

# Manage memory for large conversation histories
claude-mpm cleanup-memory

# Check for updates (including Claude Code compatibility)
claude-mpm doctor --checks updates

💡 Update Checking: Claude MPM automatically checks for updates and verifies Claude Code compatibility on startup. Configure in ~/.claude-mpm/configuration.yaml or see docs/update-checking.md.

See docs/user/getting-started.md for complete usage examples.

Managing Agent & Skill Repositories

Claude MPM uses Git repositories to distribute and manage agents and skills. This provides automatic updates, version control, and easy sharing of agent/skill libraries.

Quick Start

# Add custom agent repository
claude-mpm agent-source add https://github.com/yourorg/your-agents

# Add custom skill repository
claude-mpm skill-source add https://github.com/yourorg/your-skills

# Test repository without saving (fail-fast validation)
claude-mpm agent-source add https://github.com/yourorg/your-agents --test

# List configured sources
claude-mpm agent-source list
claude-mpm skill-source list

# Update from all sources
claude-mpm agent-source update
claude-mpm skill-source update

Priority System

When multiple sources provide the same agent/skill, priority controls which one wins:

Lower number = Higher precedence (priority 10 beats priority 100)
Local agents/skills (.claude-mpm/agents/, .claude-mpm/skills/) always override
System defaults provide fallback

Example Configuration:

# ~/.claude-mpm/config/agent_sources.yaml
repositories:
  - url: https://github.com/myteam/agents
    priority: 10    # Highest precedence (custom agents)

  - url: https://github.com/bobmatnyc/claude-mpm-agents
    priority: 100   # System default (official agents)

Hierarchical Organization with BASE-AGENT.md

Support for BASE-AGENT.md enables powerful template inheritance:

your-agents/
  BASE-AGENT.md              # Root shared content (all agents inherit)
  engineering/
    BASE-AGENT.md            # Engineering-specific (engineering/* inherit)
    python/
      fastapi-engineer.md    # Inherits both BASE files
      django-engineer.md     # Inherits both BASE files
    rust/
      systems-engineer.md    # Inherits both BASE files

How It Works:

Cascading Inheritance: Each agent inherits from all BASE-AGENT.md files in parent directories
Automatic Flattening: Nested repositories are automatically flattened during deployment
DRY Templates: Share common instructions across related agents
Maintainability: Update shared content in one place

See docs/features/hierarchical-base-agents.md for complete guide.

Configuration via YAML

Edit configuration files directly:

Agents: ~/.claude-mpm/config/agent_sources.yaml
Skills: ~/.claude-mpm/config/skill_sources.yaml

Example agent_sources.yaml:

repositories:
  - url: https://github.com/bobmatnyc/claude-mpm-agents
    priority: 100
    enabled: true

  - url: https://github.com/yourorg/custom-agents
    priority: 10
    enabled: true

Benefits

✅ Always Up-to-Date: Repositories sync automatically on startup ✅ Bandwidth Efficient: ETag-based caching reduces network usage by 95%+ ✅ Version Control: Track changes through Git history ✅ Immediate Testing: Fail-fast validation prevents startup issues ✅ Nested Support: Automatic flattening of directory structures ✅ Template Inheritance: BASE-AGENT.md for DRY principles ✅ Progress Visibility: Two-phase progress bars (sync + deployment) ✅ Community Access: Latest improvements available immediately ✅ Organization Libraries: Share team-specific content across projects

Documentation

Agent Sources User Guide - Complete guide with examples
CLI Reference - All agent-source commands
Hierarchical BASE-AGENT.md - Template inheritance guide
Migration Guide - Upgrading from v4.4.x
Troubleshooting - Common issues and solutions

Skills Deployment

Skills are automatically deployed from Git repositories, just like agents. Claude MPM includes hundreds of skills from curated repositories:

System Skills: Community skills from bobmatnyc/claude-mpm-skills
Official Skills: Anthropic's official skills from anthropics/skills
Custom Skills: Add your own skill repositories

Quick Usage

# Skills are automatically deployed on first run
# No manual deployment needed!

# Add custom skill repository
claude-mpm skill-source add https://github.com/yourorg/custom-skills

# List configured skill sources
claude-mpm skill-source list

# Update skills from all sources
claude-mpm skill-source update

How Skills Work

Automatic Deployment: All skills from configured repositories deploy on startup
Three-Tier Organization:
- Bundled: System and official skills (hundreds)
- User: Custom skills in ~/.config/claude-mpm/skills/
- Project: Project-specific skills in .claude-mpm/skills/
Priority Resolution: Project → User → Bundled (local always wins)
Auto-Linking: Skills automatically linked to relevant agents
Version Tracking: All skills support semantic versioning

Technology → Skills Mapping

The Research agent automatically analyzes your project and recommends relevant skills:

Your Technology	Auto-Recommended Skills
Python + pytest	test-driven-development, python-style
FastAPI/Flask	backend-engineer
React/Next.js	frontend-development, web-frameworks
Docker	docker-workflow
GitHub Actions	ci-cd-pipeline-builder
Playwright	webapp-testing

Interactive Management

# Interactive skills wizard
claude-mpm configure
# Choose option 2: Skills Management

# Features:
# - View all deployed skills
# - Configure skill assignments to agents
# - Auto-link skills based on agent roles
# - Manage custom skill repositories

Important Notes

⚠️ Skills load at Claude Code startup only - restart required after adding repositories
✅ Batch add related skill repositories to minimize restarts
✅ Research agent automatically recommends missing skills during project analysis
See Skills Deployment Guide for comprehensive details
See Skills Quick Reference for command reference

Architecture

Claude MPM features a clean, service-oriented architecture:

Streamlined Rich Interface: Removed complex TUI system (~2,500 lines) for cleaner user experience
MCP Integration: Full support for Model Context Protocol services with automatic detection
Service-Oriented Architecture: Simplified five specialized service domains
Interface-Based Contracts: All services implement explicit interfaces
Enhanced Performance: ~3,700 lines removed for better startup time and maintainability
Enhanced Security: Comprehensive input validation and sanitization framework

See docs/ARCHITECTURE.md for detailed architecture information.

Key Capabilities

Multi-Agent Orchestration

Claude MPM includes 15 specialized agents:

Core Development

Engineer - Software development and implementation
Research - Code analysis and research
Documentation - Documentation creation and maintenance
QA - Testing and quality assurance
Security - Security analysis and implementation

Language-Specific Engineers

Python Engineer (v2.3.0) - Type-safe, async-first Python with SOA patterns for non-trivial applications
- Service-oriented architecture with ABC interfaces for applications
- Lightweight script patterns for automation and one-off tasks
- Clear decision criteria for when to use DI/SOA vs simple functions
- Dependency injection containers with auto-resolution
- Use for: Web applications, microservices, data pipelines (DI/SOA) or scripts, CLI tools, notebooks (simple functions)
Rust Engineer (v1.1.0) - Memory-safe, high-performance systems with trait-based service architecture
- Dependency injection with traits (constructor injection, trait objects)
- Service-oriented architecture patterns (repository, builder)
- Decision criteria for when to use DI/SOA vs simple code
- Async programming with tokio and zero-cost abstractions
- Use for: Web services, microservices (DI/SOA) or CLI tools, scripts (simple code)

Operations & Infrastructure

Ops - Operations and deployment with advanced git commit authority and security verification (v2.2.2+)
Version Control - Git and version management
Data Engineer - Data pipeline and ETL development

Web Development

Web UI - Frontend and UI development
Web QA - Web testing and E2E validation

Project Management

Ticketing - Issue tracking and management
Project Organizer - File organization and structure
Memory Manager - Project memory and context management

Code Quality

Refactoring Engineer - Code refactoring and optimization
Code Analyzer - Static code analysis with AST and tree-sitter

Agent Memory System

NEW in v5.4.13: Runtime memory loading for instant updates and better efficiency.

How It Works:

Memory Files: Store agent memories in .claude-mpm/memories/{agent_id}.md (e.g., engineer.md, qa.md)
Runtime Loading: Memories loaded dynamically when agents are delegated to (no restart required)
Instant Updates: Memory changes take effect immediately
Event Observability: Memory loading tracked via EventBus (agent.memory.loaded events)

Memory Format: Agents learn project-specific patterns using a simple markdown list format and can update memories via JSON response fields (remember for incremental updates, MEMORIES for complete replacement).

See: Memory Flow Architecture for complete technical details.

Skills System

Claude MPM includes a powerful skills system that eliminates redundant agent guidance through reusable skill modules:

20 Bundled Skills covering essential development workflows (all versioned starting at 0.1.0):

Git workflow, TDD, code review, systematic debugging
API documentation, refactoring patterns, performance profiling
Docker containerization, database migrations, security scanning
JSON/PDF/XLSX handling, async testing, ImageMagick operations
Local development servers: Next.js, FastAPI, Vite, Express
Web performance: Lighthouse metrics, Core Web Vitals optimization

Three-Tier Organization:

Bundled: Core skills included with Claude MPM (~15,000 lines of reusable guidance)
User: Custom skills in ~/.config/claude-mpm/skills/
Project: Project-specific skills in .claude-mpm/skills/

Version Tracking:

All skills support semantic versioning (MAJOR.MINOR.PATCH)
Check versions with /mpm-version command in Claude Code
See Skills Versioning Guide for details

Quick Access:

# Interactive skills management
claude-mpm configure
# Choose option 2: Skills Management

# Auto-link skills to agents based on their roles
# Configure custom skill assignments
# View current skill mappings

Skills are automatically injected into agent prompts, reducing template size by 85% while maintaining full capability coverage.

MCP Gateway (Model Context Protocol)

Claude MPM includes a powerful MCP Gateway that enables:

Integration with external tools and services
Custom tool development
Protocol-based communication
Extensible architecture

See MCP Gateway Documentation for details.

Memory Management

Large conversation histories can consume 2GB+ of memory. Use the cleanup-memory command to manage Claude conversation history:

# Clean up old conversation history
claude-mpm cleanup-memory

# Keep only recent conversations
claude-mpm cleanup-memory --days 7

Resume Log System

NEW in v4.17.2 - Proactive context management for seamless session continuity.

The Resume Log System automatically generates structured 10k-token logs when approaching Claude's context window limits, enabling you to resume work without losing important context.

Key Features:

🎯 Graduated Thresholds: Warnings at 70% (60k buffer), 85% (30k buffer), and 95% (10k buffer)
📋 Structured Logs: 10k-token budget intelligently distributed across 7 key sections
🔄 Seamless Resumption: Automatically loads previous session context on startup
📁 Human-Readable: Markdown format for both Claude and human review
⚙️ Zero-Configuration: Works automatically with sensible defaults

How It Works:

Monitor token usage continuously throughout session
Display proactive warnings at 70%, 85%, and 95% thresholds
Automatically generate resume log when approaching limits
Load previous resume log when starting new session
Continue work seamlessly with full context preservation

Example Resume Log Structure:

# Session Resume Log: 20251101_115000

## Context Metrics (500 tokens)
- Token usage and percentage

## Mission Summary (1,000 tokens)
- Overall goal and purpose

## Accomplishments (2,000 tokens)
- What was completed

## Key Findings (2,500 tokens)
- Important discoveries

## Decisions & Rationale (1,500 tokens)
- Why choices were made

## Next Steps (1,500 tokens)
- What to do next

## Critical Context (1,000 tokens)
- Essential state, IDs, paths

Configuration (.claude-mpm/configuration.yaml):

context_management:
  enabled: true
  budget_total: 200000
  thresholds:
    caution: 0.70   # First warning - plan transition
    warning: 0.85   # Strong warning - wrap up
    critical: 0.95  # Urgent - stop new work
  resume_logs:
    enabled: true
    auto_generate: true
    max_tokens: 10000
    storage_dir: ".claude-mpm/resume-logs"

QA Status: 40/41 tests passing (97.6% coverage), APPROVED FOR PRODUCTION ✅

See docs/user/resume-logs.md for complete documentation.

Real-Time Monitoring

The --monitor flag opens a web dashboard showing live agent activity, file operations, and session management.

See docs/reference/MEMORY.md and docs/developer/11-dashboard/README.md for details.

📚 Documentation

👉 Complete Documentation Hub - Start here for all documentation!

Quick Links by User Type

👥 For Users

🚀 5-Minute Quick Start - Get running immediately
📦 Installation Guide - All installation methods
📖 User Guide - Complete user documentation
❓ FAQ - Common questions answered

💻 For Developers

🏗️ Architecture Overview - Service-oriented system design
💻 Developer Guide - Complete development documentation
🧪 Contributing - How to contribute
📊 API Reference - Complete API documentation

🤖 For Agent Creators

🤖 Agent System - Complete agent development guide
📝 Creation Guide - Step-by-step tutorials
📋 Schema Reference - Agent format specifications

🚀 For Operations

🚀 Deployment - Release management & versioning
📊 Monitoring - Real-time dashboard & metrics
🐛 Troubleshooting - Enhanced doctor command with detailed reports and auto-fix capabilities

🎯 Documentation Features

Single Entry Point: docs/README.md is your navigation hub
Clear User Paths: Organized by user type and experience level
Cross-Referenced: Links between related topics and sections
Up-to-Date: Version 4.16.3 with web performance optimization skill

Recent Updates (v5.0) 🎉

Major Release: Git Repository Integration for Agents & Skills

🚀 What's New:

Git Repository Support: Agents and skills now deploy from Git repositories
47+ Agents: Comprehensive agent library from curated repositories
Hundreds of Skills: System skills + Official Anthropic skills automatically deployed
Hierarchical BASE-AGENT.md: Template inheritance for DRY principles
Nested Repository Support: Automatic flattening of directory structures
Immediate Testing: Fail-fast validation when adding repositories
Two-Phase Progress: Clear visibility during sync and deployment
ETag Caching: 95%+ bandwidth reduction with intelligent caching

🔧 Repository Management:

claude-mpm agent-source add/list/update - Manage agent repositories
claude-mpm skill-source add/list/update - Manage skill repositories
Priority-based resolution for multiple sources
YAML configuration support

📚 Documentation:

Complete guides for Git repository integration
Hierarchical BASE-AGENT.md documentation
Migration guides for existing installations
Troubleshooting and best practices

🎯 Benefits:

Always up-to-date content from repositories
Community-driven agent and skill libraries
Easy sharing of organizational content
Version control for all templates

See CHANGELOG.md for full history and docs/user/MIGRATION.md for upgrade instructions.

📜 License

Licensed under the Elastic License 2.0 - free for internal use and commercial products.

Main restriction: Cannot offer as a hosted SaaS service without a commercial license.

📖 Licensing FAQ | 💼 Commercial licensing: bob@matsuoka.com

Development

Quick Development Setup

# Complete development setup with code formatting and quality tools
make dev-complete

# Or step by step:
make setup-dev          # Install in development mode
make setup-pre-commit    # Set up automated code formatting

Code Quality & Formatting

The project uses automated code formatting and quality checks:

Black for code formatting
isort for import sorting
flake8 for linting
mypy for type checking
Pre-commit hooks for automatic enforcement

See docs/developer/CODE_FORMATTING.md for details.

Contributing

Contributions are welcome! Please see our project structure guide and follow the established patterns.

Development Workflow:

Run make dev-complete to set up your environment
Code formatting happens automatically on commit
All code must pass quality checks before merging

Project Structure

See docs/reference/STRUCTURE.md for codebase organization.

Credits

Based on claude-multiagent-pm
Enhanced for Claude Code (CLI) integration
Built with ❤️ by the Claude MPM community

Name		Name	Last commit message	Last commit date
Latest commit History 1,750 Commits
.githooks		.githooks
.github/workflows		.github/workflows
_archive		_archive
config		config
docker		docker
docs		docs
examples		examples
kuzu-memories		kuzu-memories
npm-bin		npm-bin
scripts		scripts
src/claude_mpm		src/claude_mpm
tests		tests
tickets		tickets
tools		tools
.flake8		.flake8
.gitignore		.gitignore
.pre-commit-config.yaml		.pre-commit-config.yaml
.python-version		.python-version
.readthedocs.yaml		.readthedocs.yaml
.ruff.toml		.ruff.toml
.secrets.baseline		.secrets.baseline
.trackdown.yaml		.trackdown.yaml
BUILD_NUMBER		BUILD_NUMBER
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
CONTRIBUTING.md		CONTRIBUTING.md
DASHBOARD_FILE_VIEWER_FIX.md		DASHBOARD_FILE_VIEWER_FIX.md
DEBUG-EVENT-LOGGING.md		DEBUG-EVENT-LOGGING.md
DIAGNOSIS.md		DIAGNOSIS.md
EVENT_SCHEMA_COMPARISON.md		EVENT_SCHEMA_COMPARISON.md
FINDINGS.md		FINDINGS.md
FIX_SUMMARY_PM_INSTRUCTIONS_VERSION_VALIDATION.md		FIX_SUMMARY_PM_INSTRUCTIONS_VERSION_VALIDATION.md
HOOK_FIX_COMMANDS.md		HOOK_FIX_COMMANDS.md
LICENSE		LICENSE
LICENSE-FAQ.md		LICENSE-FAQ.md
MPM_EVENT_SCHEMA_FIX.md		MPM_EVENT_SCHEMA_FIX.md
Makefile		Makefile
README.md		README.md
SKILL_PARSING_FIX.md		SKILL_PARSING_FIX.md
VERSION		VERSION
VERSION_POLICY.md		VERSION_POLICY.md
claude-mpm-dev		claude-mpm-dev
inspect-tool-event.js		inspect-tool-event.js
package.json		package.json
pyproject.toml		pyproject.toml
test-files-tab-verification.js		test-files-tab-verification.js
uv.lock		uv.lock
verify-files-tab.js		verify-files-tab.js
vite.config.js		vite.config.js

License

PeterBell/claude-mpm

Folders and files

Latest commit

History

Repository files navigation

Claude MPM - Multi-Agent Project Manager

🚀 What's New in v5.0

Git Repository Integration for Agents & Skills

Quick Start: Add Your Own Repository

Default Repositories

Hierarchical Organization Example

Features

🎯 Multi-Agent Orchestration

📦 Git Repository Integration (NEW in v5.0)

🎯 Skills System

🔌 Advanced Integration

⚡ Performance & Security

Quick Installation

Prerequisites

Install Claude MPM

Verify Installation

🤝 Recommended Partner Products

kuzu-memory - Advanced Memory Management

mcp-vector-search - Semantic Code Search

Quick Setup - Both Tools

Cache Management

Cache Structure

Git Workflow (Optional)

HTTP Sync Fallback

Migration from Legacy Cache

Quick Usage

Managing Agent & Skill Repositories

Quick Start

Priority System

Hierarchical Organization with BASE-AGENT.md

Configuration via YAML

Benefits

Documentation

Skills Deployment

Quick Usage

How Skills Work

Technology → Skills Mapping

Interactive Management

Important Notes

Architecture

Key Capabilities

Multi-Agent Orchestration

Core Development

Language-Specific Engineers

Operations & Infrastructure

Web Development

Project Management

Code Quality

Agent Memory System

Skills System

MCP Gateway (Model Context Protocol)

Memory Management

Resume Log System

Real-Time Monitoring

📚 Documentation

Quick Links by User Type

👥 For Users

💻 For Developers

🤖 For Agent Creators

🚀 For Operations

🎯 Documentation Features

Recent Updates (v5.0) 🎉

📜 License

Development

Quick Development Setup

Code Quality & Formatting

Contributing

Project Structure

Credits

About

Resources

License

Contributing

Uh oh!

Packages