Home
Claude-Flow v2 Alpha is an enterprise-grade AI orchestration platform that revolutionizes how developers build with AI. By combining hive-mind swarm intelligence, neural pattern recognition, and 87 advanced MCP tools, Claude-Flow enables unprecedented AI-powered development workflows.
- 🔍 NEW: Truth Verification System: Mandatory verification with 0.95 accuracy threshold
- 👥 NEW: Pair Programming Mode: Real-time collaborative development with continuous validation
- 🧠 NEW: Training Pipeline: Machine learning system that improves agent performance over time
- 🤖 64 Specialized AI Agents: Complete agent ecosystem across 16 categories
- 🐝 Hive-Mind Intelligence: Queen-led AI coordination with specialized worker agents
- ✨ NEW: Intelligent Automation: MLE-STAR workflow engine with smart agent spawning
- 🔗 NEW: Stream-JSON Chaining: Real-time agent-to-agent output piping for seamless coordination
- 🧠 Neural Networks: 27+ cognitive models with WASM SIMD acceleration
- 🔧 87 MCP Tools: Comprehensive toolkit for swarm orchestration, memory, and automation
- 🔄 Dynamic Agent Architecture (DAA): Self-organizing agents with fault tolerance
-
💾 SQLite Memory System: Persistent
.swarm/memory.dbwith 12 specialized tables - 🪝 Advanced Hooks System: Automated workflows with pre/post operation hooks
- 📊 GitHub Integration: 13 specialized agents for repository management
- 🎉 NEW: GitHub-Enhanced Init: Automatic GitHub releases for all checkpoints
- 📈 NEW: Benchmark System: Comprehensive performance testing with swarm-bench CLI
- ⚡ Performance: 84.8% SWE-Bench solve rate, 2.8-4.4x speed improvement
Initialize powerful AI modules with a single command:
# Initialize SAFLA Neural Module - Self-learning AI systems
npx claude-flow@alpha neural init
npx claude-flow@alpha neural init --force # Overwrite existing
# Initialize GOAP Goal Module - Intelligent planning systems
npx claude-flow@alpha goal init
npx claude-flow@alpha goal init --force # Overwrite existing- Neural Module: SAFLA self-aware systems with 4-tier memory, 172K+ ops/sec
- Goal Module: GOAP planning with A* pathfinding and adaptive replanning
- Ready-to-use agents for Claude Code (
@agent-safla-neural,@agent-goal-planner) - Complete configuration and documentation
Claude-Flow now enforces "truth is enforced, not assumed" with mandatory verification:
# Initialize project with verification
npx claude-flow@alpha init --verify # Truth verification mode
npx claude-flow@alpha init --pair # Pair programming mode
npx claude-flow@alpha init --verify --pair # Combined mode
# Run verification system
npx claude-flow@alpha verify init strict # 0.95 threshold, auto-rollback
npx claude-flow@alpha verify status # Check system status
# Run verification
npx claude-flow@alpha verify verify task-123 --agent coder
npx claude-flow@alpha truth # View truth scores
# Pair programming with real-time verification
npx claude-flow@alpha pair --start # Start collaborative sessionKey Verification Features:
- ✅ 0.95 Truth Threshold: 95% accuracy required for production
- 🔄 Auto-Rollback: Automatic recovery from verification failures
- 🔒 Byzantine Fault Tolerance: Protection against incorrect agents
- 📊 Real-time Monitoring: Continuous validation and feedback
- 🤝 Pair Programming: Human + AI collaborative development
📚 Truth Verification System Documentation → 📚 Pair Programming System Documentation →
Claude-Flow Alpha.88 includes a comprehensive agent system with 64 specialized AI agents across 16 categories, providing unparalleled development capabilities:
- Agent System Overview - Complete catalog of all 64 agents with descriptions
- Agent Usage Guide - Practical examples and concurrent usage patterns
- Agent Categories - Detailed breakdown of all 16 categories
- Core Development (5): coder, planner, researcher, reviewer, tester
- Swarm Coordination (3): hierarchical, mesh, adaptive coordinators
- Consensus Systems (7): Byzantine, Raft, Gossip, CRDT protocols
- GitHub Integration (13): PR management, code review, release automation
- Performance Optimization (6): monitoring, load balancing, topology optimization
- View All 16 Categories →
# Initialize with complete agent system
npx claude-flow@alpha init
# Deploy multi-agent swarm
claude-flow swarm "Build REST API with authentication" --agents 8
# Concurrent agent deployment
Task("System architecture", "...", "system-architect")
Task("Backend development", "...", "backend-dev")
Task("Test creation", "...", "tester")Claude-Flow provides specialized CLAUDE.md templates for different project types and development patterns. These templates optimize AI coordination for specific use cases:
| 🏷️ Category | 📊 Templates | 🎯 Use Cases |
|---|---|---|
| 🚀 Project Types | Web Dev, Mobile, API, AI/ML | Full-stack apps, mobile apps, APIs, ML projects |
| 🧪 Methodologies | TDD, Agile, DDD, CI/CD | Test-driven, sprint-based, domain-focused development |
| 🏗️ Architecture | Microservices, Serverless, Monolithic | Distributed systems, cloud-native, traditional apps |
| 🔧 Languages | JavaScript, Python, TypeScript, Java | Language-specific patterns and best practices |
| 🏢 Team Size | Solo, Small Team, Enterprise | Individual, startup, corporate development |
💡 Quick Start:
claude-flow templates apply web-development --output CLAUDE.md
➡️ Browse All Templates - Complete collection with usage guides
- Neural Module - SAFLA self-learning systems with persistent memory
- Goal Module - GOAP intelligent planning with A* pathfinding
- Installation Guide - Complete setup instructions
- Quick Start - Get up and running in minutes
- Init Commands - Project initialization options and configurations
- Windows Installation - Windows-specific setup
- Non-Interactive Mode - CI/CD and automation guide
- GitHub Actions Integration - Automated workflows
- WebSocket Server Guide - Headless WebSocket deployment
- Hive-Mind Intelligence - Understanding AI coordination
- Agent System Overview - Complete 64-agent system reference
- Agent Usage Guide - Practical usage patterns and examples
- Memory System - SQLite-based persistent memory
- Hooks System - Automated workflow orchestration
- Truth Verification System 🔍 - Mandatory verification with 0.95 truth threshold
- Pair Programming System 👥 - Real-time collaborative development with AI
- Training Pipeline (NEW!) 🧠 - Machine learning system for continuous agent improvement
- Verification-Training Integration 🔄 - How verification feeds training for better predictions
- Neural Networks - AI pattern recognition and learning
- MCP Tools - Complete tool reference (87 tools)
- GitHub Integration - Repository management automation
- GitHub Hooks (NEW!) - Automatic checkpoint releases & team collaboration
- Workflow Orchestration - Complex task coordination
- Automation Commands (NEW!) - Intelligent workflow automation with MLE-STAR
- Stream Chaining (NEW!) - Real-time agent output piping and coordination with technical implementation details
- MLE-STAR Workflow (NEW!) - Machine Learning Engineering via Search and Targeted Refinement
- Benchmark System - Complete benchmarking framework with swarm-bench CLI
- SWE-bench Evaluation (NEW!) - Official software engineering benchmark integration
- Performance Benchmarking (NEW!) - Advanced performance testing and optimization
- SPARC Methodology - Test-driven development patterns
- Development Patterns - Best practices and examples
- API Reference - Complete API documentation
- Troubleshooting - Common issues and solutions
- CLAUDE.md Templates - Project-specific configuration templates
- Web Development - Full-stack web application template
- TDD Template - Test-driven development configuration
- API Development - RESTful API project template
- Mobile Development - Native and hybrid mobile apps
- Contributing - How to contribute to Claude-Flow
- Examples - Real-world usage examples
- FAQ - Frequently asked questions
# Install Claude-Flow Alpha
npm install -g claude-flow@alpha
# Standard Init (Local Git checkpoints)
claude-flow init
# GitHub-Enhanced Init (NEW! GitHub releases for checkpoints)
claude-flow github init
# Initialize Hive-Mind Swarm
claude-flow hive init --topology mesh --agents 5
# Run SPARC Development Mode
claude-flow sparc run dev "build user authentication"
# Execute with full orchestration
claude-flow orchestrate "create REST API with tests" --agents 8 --parallel
# NEW! Intelligent Automation Commands
claude-flow automation mle-star --dataset data.csv --target label --claude
claude-flow automation auto-agent --task-complexity enterprise
claude-flow automation run-workflow workflow.json --claude --non-interactive
# NEW! Stream Chaining (Agent-to-Agent Piping)
claude-flow automation mle-star --dataset data.csv --target price --claude --output-format stream-json
# NEW! Training Pipeline Commands
claude-flow train-pipeline run --complexity medium --iterations 3
claude-flow train-pipeline status # View agent performance profiles
claude-flow train-pipeline validate # Check current performance metrics
claude-flow verify-train feed # Feed verification data to training
claude-flow verify-train predict default coder # Predict task outcomes
# NEW! Benchmarking Commands
swarm-bench run "Build REST API" --strategy development --max-agents 6
swarm-bench swarm "Create auth system" --mode hierarchical --parallel
swarm-bench sparc coder "Implement feature" --timeout 60
# NEW! SWE-bench Integration (Official software engineering benchmark)
swarm-bench swe-bench official --limit 10 # Test with 10 instances
swarm-bench swe-bench multi-mode --instances 5 # Compare all modes
swarm-bench swe-bench official --lite # Full evaluation (300 instances)🔥 Revolutionary AI Coordination: Build faster, smarter, and more efficiently with AI-powered development orchestration
Latest Version: v2.0.0-alpha.88 | License: MIT | Node.js: 18+ Required