@tom_doerr: Runs Claude agents 24/7 inside tmux https://github.com/Jedward23/Tmux-Orchestrator…
Summary
Tmux-Orchestrator is an open-source tool that enables autonomous 24/7 operation of Claude AI agents using a multi-agent hierarchy within tmux sessions for parallel project management and coding.
View Cached Full Text
Cached at: 05/10/26, 08:31 PM
Runs Claude agents 24/7 inside tmux
https://t.co/hgH0awBoRY https://t.co/YIROMvZ9sp
Jedward23/Tmux-Orchestrator
Source: https://github.com/Jedward23/Tmux-Orchestrator

Run AI agents 24/7 while you sleep - The Tmux Orchestrator enables Claude agents to work autonomously, schedule their own check-ins, and coordinate across multiple projects without human intervention.
🤖 Key Capabilities & Autonomous Features
- Self-trigger - Agents schedule their own check-ins and continue work autonomously
- Coordinate - Project managers assign tasks to engineers across multiple codebases
- Persist - Work continues even when you close your laptop
- Scale - Run multiple teams working on different projects simultaneously
🏗️ Architecture
The Tmux Orchestrator uses a three-tier hierarchy to overcome context window limitations:
┌─────────────┐
│ Orchestrator│ ← You interact here
└──────┬──────┘
│ Monitors & coordinates
▼
┌─────────────┐ ┌─────────────┐
│ Project │ │ Project │
│ Manager 1 │ │ Manager 2 │ ← Assign tasks, enforce specs
└──────┬──────┘ └──────┬──────┘
│ │
▼ ▼
┌─────────────┐ ┌─────────────┐
│ Engineer 1 │ │ Engineer 2 │ ← Write code, fix bugs
└─────────────┘ └─────────────┘
Why Separate Agents?
- Limited context windows - Each agent stays focused on its role
- Specialized expertise - PMs manage, engineers code
- Parallel work - Multiple engineers can work simultaneously
- Better memory - Smaller contexts mean better recall
📸 Examples in Action
Project Manager Coordination
The orchestrator creating and briefing a new project manager agent
Status Reports & Monitoring
Real-time status updates from multiple agents working in parallel
Tmux Communication
How agents communicate across tmux windows and sessions
Project Completion
Successful project completion with all tasks verified and committed
🎯 Quick Start
Option 1: Basic Setup (Single Project)
# 1. Create a project spec
cat > project_spec.md << 'EOF'
PROJECT: My Web App
GOAL: Add user authentication system
CONSTRAINTS:
- Use existing database schema
- Follow current code patterns
- Commit every 30 minutes
- Write tests for new features
DELIVERABLES:
1. Login/logout endpoints
2. User session management
3. Protected route middleware
EOF
# 2. Start tmux session
tmux new-session -s my-project
# 3. Start project manager in window 0
claude
# 4. Give PM the spec and let it create an engineer
"You are a Project Manager. Read project_spec.md and create an engineer
in window 1 to implement it. Schedule check-ins every 30 minutes."
# 5. Schedule orchestrator check-in
./schedule_with_note.sh 30 "Check PM progress on auth system"
Option 2: Full Orchestrator Setup
# Start the orchestrator
tmux new-session -s orchestrator
claude
# Give it your projects
"You are the Orchestrator. Set up project managers for:
1. Frontend (React app) - Add dashboard charts
2. Backend (FastAPI) - Optimize database queries
Schedule yourself to check in every hour."
✨ Key Features
🔄 Self-Scheduling Agents
Agents can schedule their own check-ins using:
./schedule_with_note.sh 30 "Continue dashboard implementation"
👥 Multi-Agent Coordination
- Project managers communicate with engineers
- Orchestrator monitors all project managers
- Cross-project knowledge sharing
💾 Automatic Git Backups
- Commits every 30 minutes of work
- Tags stable versions
- Creates feature branches for experiments
📊 Real-Time Monitoring
- See what every agent is doing
- Intervene when needed
- Review progress across all projects
📋 Best Practices
Writing Effective Specifications
PROJECT: E-commerce Checkout
GOAL: Implement multi-step checkout process
CONSTRAINTS:
- Use existing cart state management
- Follow current design system
- Maximum 3 API endpoints
- Commit after each step completion
DELIVERABLES:
1. Shipping address form with validation
2. Payment method selection (Stripe integration)
3. Order review and confirmation page
4. Success/failure handling
SUCCESS CRITERIA:
- All forms validate properly
- Payment processes without errors
- Order data persists to database
- Emails send on completion
Git Safety Rules
-
Before Starting Any Task
git checkout -b feature/[task-name] git status # Ensure clean state -
Every 30 Minutes
git add -A git commit -m "Progress: [what was accomplished]" -
When Task Completes
git tag stable-[feature]-[date] git checkout main git merge feature/[task-name]
🚨 Common Pitfalls & Solutions
| Pitfall | Consequence | Solution |
|---|---|---|
| Vague instructions | Agent drift, wasted compute | Write clear, specific specs |
| No git commits | Lost work, frustrated devs | Enforce 30-minute commit rule |
| Too many tasks | Context overload, confusion | One task per agent at a time |
| No specifications | Unpredictable results | Always start with written spec |
| Missing checkpoints | Agents stop working | Schedule regular check-ins |
🛠️ How It Works
The Magic of Tmux
Tmux (terminal multiplexer) is the key enabler because:
- It persists terminal sessions even when disconnected
- Allows multiple windows/panes in one session
- Claude runs in the terminal, so it can control other Claude instances
- Commands can be sent programmatically to any window
💬 Simplified Agent Communication
We now use the send-claude-message.sh script for all agent communication:
# Send message to any Claude agent
./send-claude-message.sh session:window "Your message here"
# Examples:
./send-claude-message.sh frontend:0 "What's your progress on the login form?"
./send-claude-message.sh backend:1 "The API endpoint /api/users is returning 404"
./send-claude-message.sh project-manager:0 "Please coordinate with the QA team"
The script handles all timing complexities automatically, making agent communication reliable and consistent.
Scheduling Check-ins
# Schedule with specific, actionable notes
./schedule_with_note.sh 30 "Review auth implementation, assign next task"
./schedule_with_note.sh 60 "Check test coverage, merge if passing"
./schedule_with_note.sh 120 "Full system check, rotate tasks if needed"
Important: The orchestrator needs to know which tmux window it’s running in to schedule its own check-ins correctly. If scheduling isn’t working, verify the orchestrator knows its current window with:
echo "Current window: $(tmux display-message -p "#{session_name}:#{window_index}")"
🎓 Advanced Usage
Multi-Project Orchestration
# Start orchestrator
tmux new-session -s orchestrator
# Create project managers for each project
tmux new-window -n frontend-pm
tmux new-window -n backend-pm
tmux new-window -n mobile-pm
# Each PM manages their own engineers
# Orchestrator coordinates between PMs
Cross-Project Intelligence
The orchestrator can share insights between projects:
- “Frontend is using /api/v2/users, update backend accordingly”
- “Authentication is working in Project A, use same pattern in Project B”
- “Performance issue found in shared library, fix across all projects”
📚 Core Files
send-claude-message.sh- Simplified agent communication scriptschedule_with_note.sh- Self-scheduling functionalitytmux_utils.py- Tmux interaction utilitiesCLAUDE.md- Agent behavior instructionsLEARNINGS.md- Accumulated knowledge base
🤝 Contributing & Optimization
The orchestrator evolves through community discoveries and optimizations. When contributing:
- Document new tmux commands and patterns in CLAUDE.md
- Share novel use cases and agent coordination strategies
- Submit optimizations for claudes synchronization
- Keep command reference up-to-date with latest findings
- Test improvements across multiple sessions and scenarios
Key areas for enhancement:
- Agent communication patterns
- Cross-project coordination
- Novel automation workflows
📄 License
MIT License - Use freely but wisely. Remember: with great automation comes great responsibility.
“The tools we build today will program themselves tomorrow” - Alan Kay, 1971
Similar Articles
@tom_doerr: Claude Code scaffolding with multi-agent teams and TDD https://github.com/alinaqi/claude-bootstrap…
Maggy v5.0 is an open-source CLI framework that enhances Claude Code into an autonomous engineering platform with multi-agent orchestration, TDD loops, and built-in security protocols. It enables structured workflows through containerized agent teams, cross-machine sync, and automated quality gates.
@tom_doerr: Orchestrates AI agents in isolated git worktrees https://github.com/txtx/axel-app
Axel is a macOS app with a Rust CLI that orchestrates AI agents in isolated git worktrees and tmux panes, providing a task management interface for multi-agent workflows.
@tom_doerr: Automates sales and marketing tasks in Claude Code https://github.com/gtmagents/gtm-agents…
An open-source tool that automates sales, marketing, and growth tasks using Claude Code, saving teams 15+ hours per week on repetitive busywork with no coding required.
@tom_doerr: Runs a virtual company with 14 expert AI agents https://github.com/MaxMiksa/Auto-Company…
Auto-Company is an open-source project that orchestrates 14 autonomous AI agents to run a fully automated company 24/7, handling ideation, coding, deployment, and marketing without human intervention, powered by Claude Code and Codex CLI.
@JinjingLiang: Want to ditch Claude? Someone just released the any-agent version of 'Claude Agent View' Use Codex, Pi, Droid, whatever…
Orca is an open-source AI orchestrator that lets developers run multiple CLI agents like Claude Code, Codex, and Gemini side-by-side across repositories on macOS, Windows, and Linux.