"Stay awhile and listen... Transform one Claude into an infinite swarm."
Run 10+ specialized agents in parallel. Build at the speed of thought.
You've been using Claude Code like a single warrior. This system awakens the swarm:
- Engineers building different modules simultaneously
- Architects designing while others implement
- Researchers gathering wisdom as code flows
- Gatekeepers ensuring perfection in parallel
- Recursive Orchestration: Infinite orchestrators managing infinite squads
Watch your legion work through the real-time dashboard. Feel the power.
Context-engineered for extended battles without bankruptcy:
- Minimal context per agent (they share memory through SQLite)
- File references, not contents (massive token savings)
- From the Maker: I rarely hit my claude usage cap, even with 3-5 concurrent sessions running all day.
# 1. Clone this repo and start the system
./scripts/start-system.sh
# If this fails you probably need to bun install or whatever. Just ask claude code to sort it out.
# 2. Use it in your project
npx claude-comms # in your project root.
# or copy CLAUDE.md and .claude directory to your projects
# 3. Launch claude code in your project. --dangerously-skip-permissions if you like.
# 4. prompt claude to get cookin: /cook <what you want to build, requirements spec, etc>
"/cook make a dashboard that for the real time event stream but style it like the matrix and populate the matrix characters with the event stream data"-
Observe
http://localhost:5173/./scripts/restart-system.sh restarts everything and can clear the db -
Install perplexity mcp. it will make your claude code cap last 20% longer. I've used $5 in a month and I'm a HEAVY user.
claude mcp add perplexity-ask -- env PERPLEXITY_API_KEY=pplx-xxxx npx -y server-perplexity-askYou can still be a wizzard without this, but its the difference between being Dumbledore and being Gandalf. -
In this project prompt claude with
/learnMy man Deckard Cain will teach you everything I know.
Claude Agents → Hook Scripts → HTTP POST → Bun Server → SQLite → WebSocket → Vue Dashboard
↓ ↓
Communication Hooks Message Routing
↓ ↓
Inter-Agent Messages Subagent Registry
Tech Stack: Vue 3, Bun, TypeScript, SQLite, WebSocket, Python hooks
- Hook Events: Pre/Post tool use, notifications, user prompts
- Real-time Capture: SQLite storage with WebSocket broadcasting
- Timeline View: Visual event stream with filtering and search
- Agent Registry: Automatic discovery and session tracking
- Message Queue: Inter-agent messaging with read receipts
- Dashboard Views: Live communication monitoring
- Copy
.claudedirectory to any project for instant observability - Configure
settings.jsonwith your project name - Start monitoring - no code changes required
| Command | Description |
|---|---|
./scripts/start-system.sh |
Launch server and dashboard |
./scripts/reset-system.sh |
Clean shutdown and reset |
./scripts/test-system.sh |
System validation |
- Setup Guide - Installation and quick start (< 5 minutes)
- Integration Guide - Adding to your projects
- Architecture Guide - System design and components
- API Reference - Complete endpoint documentation
- Design System Guide - UI components and patterns
- Orchestration Guide - Multi-agent coordination patterns
- Development Guide - Contributing and local setup
- Troubleshooting Guide - Common issues and solutions
- Legacy Documentation - Archived guides and historical technical documentation
# Observability
POST /events # Submit hook events
GET /events/recent # Retrieve event timeline
WS /stream # Real-time event stream
# Multi-Agent Communication
POST /subagents/register # Register new agent
POST /subagents/message # Send inter-agent message
POST /subagents/unread # Get unread messages# Send message between agents
uv run .claude/hooks/comms/send_message.py --sender "AgentName" --message "Hello"
# Check for unread messages
uv run .claude/hooks/comms/get_unread_messages.py --name "AgentName"# System validation
./scripts/test-system.sh
# Manual event test
curl -X POST http://localhost:4000/events \
-H "Content-Type: application/json" \
-d '{"source_app": "test", "session_id": "test-123", "hook_event_type": "PreToolUse"}'
# Multi-agent communication test
python3 .claude/hooks/comms/send_message.py --sender "Agent-Alpha" --message "Test message"- Claude Code - Anthropic's official CLI
- Astral uv - Python package manager for hooks
- Bun or npm/yarn - JavaScript runtime
- API Keys - Anthropic (required), OpenAI/ElevenLabs (optional)
# Copy environment template
cp .env.sample .env
# Configure API keys in .env
ANTHROPIC_API_KEY=your_key_here
OPENAI_API_KEY=optional_key
ELEVEN_API_KEY=optional_key- Server:
localhost:4000(HTTP/WebSocket) - Dashboard:
localhost:5173(Vite dev server)
| Event | Description | Use Case |
|---|---|---|
| PreToolUse | Before tool execution | Validation & blocking |
| PostToolUse | After tool completion | Result tracking |
| UserPromptSubmit | User prompt submission | Session analysis |
| Notification | User interactions | UX monitoring |
| SubagentStop | Agent completion | Orchestration tracking |
- Real-time Timeline: Live event stream with auto-scroll
- Advanced Filtering: By app, session, event type, time range
- Chat Transcripts: Full conversation history with AI summaries
- Agent Communications: Live inter-agent message display
- Visual Analytics: Pulse charts and session visualization
- Fork the repository
- Create a feature branch:
git checkout -b feature/amazing-feature - Follow the Contributing Guide
- Submit a pull request
MIT License - see LICENSE file for details.
- Anthropic Claude Code Team - For the fantastic hooks system
- Contributors - Community feedback and improvements
- Open Source Dependencies - Vue, Bun, and ecosystem tools
For comprehensive documentation and advanced features, explore our complete guides.