A terminal-based skill for spawning, monitoring, and managing multiple AI coding agents in persistent background tmux sessions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install tmux-agents
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install tmux-agents using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Tmux Agents?
Tmux Agents is a specialized tool designed to handle long-running or parallel coding tasks by offloading them to background tmux sessions. As a key component of Openclaw Skills, it allows developers to initiate sessions for various AI models—including Claude, OpenAI Codex, and Google Gemini—without locking up their active terminal environment.
This skill is particularly valuable for developers who need to run complex refactoring or testing tasks that might take several minutes or hours. By utilizing background sessions, you can maintain a high-velocity workflow while multiple Openclaw Skills agents work asynchronously on different parts of your codebase, providing a seamless transition between cloud-powered intelligence and local, cost-free execution via Ollama.
Tmux Agents Use Cases
- Offloading complex refactoring tasks to the background to keep the main terminal free.
- Running multiple AI agents in parallel to handle backend, frontend, and documentation tasks simultaneously.
- Conducting long-running experiments with local models via Ollama to save on API costs.
- Maintaining persistent AI coding sessions that survive terminal restarts or network disconnections.
- Monitoring the progress of automated coding tasks without active supervision.
How Tmux Agents Works
- The user executes a spawn script with a specific task name, objective, and model selection.
- The skill initializes a new, isolated tmux session in the background.
- The selected agent (cloud-based like Claude or local via Ollama) begins processing the task within that session.
- Developers use status scripts to monitor progress across all active Openclaw Skills agent sessions.
- If interaction is required, the user can attach to the specific tmux session to provide real-time input.
- Once the task is complete, the user reviews the changes and kills the session to release system resources.
Tmux Agents Setup
To use this skill, ensure that tmux is installed on your local machine. If you plan to use local agents, Ollama must also be configured.
# Install tmux using Homebrew
brew install tmux
# For local agents, pull the recommended coding model
ollama pull glm-4.7-flash
# Perform a one-time configuration for local tools
ollama launch claude --model glm-4.7-flash --config
Once the dependencies are met, you can use the Openclaw Skills scripts located in the ./skills/tmux-agents/scripts/ directory.
Tmux Agents Data Schema & Taxonomy
The skill manages task metadata and execution state through tmux session naming conventions and specific script arguments.
| Attribute | Description |
|---|---|
| Session Name | A unique identifier for the background task (e.g., 'fix-bug') |
| Agent Type | Defines the engine (claude, codex, gemini, or ollama variants) |
| Status | Current activity of the tmux session reachable via status.sh |
| Persistence | Sessions remain active in the system process tree until explicitly killed |
Tmux Agents Advanced Features
- Parallel Agent Orchestration: Trigger multiple agents across different models to work on separate modules simultaneously.
- Hybrid Cloud-Local Workflows: Assign production-critical tasks to cloud models and experimental refactors to local Ollama models.
- Live Session Attachment: Seamlessly jump into a background process to watch the agent code in real-time.
- Persistent Lifecycle: Ensures that background coding tasks continue even if the primary dev environment is closed.
- Integrated Tooling: Built-in support for Claude Code and OpenAI Codex within the Openclaw Skills framework.
SKILL.md
Loading
Related Openclaw Skills
Health Guardian
A proactive health intelligence skill for AI agents that monitors Apple Health data to detect patterns and alert on vital anomalies.
Health Guardian
A proactive health intelligence skill for AI agents that monitors Apple Health data to detect patterns and alert on physiological anomalies.
Health Guardian
A proactive health intelligence skill for AI agents that monitors Apple Health metrics to detect anomalies and manage chronic conditions.
Accessibility Toolkit
A comprehensive framework for AI agents designed to eliminate interaction friction for users with physical disabilities through proactive automation and voice-first patterns.
Memoria System
A comprehensive long-term memory management system that provides AI assistants with human-like cognitive memory layers.
NEAR Tools
A comprehensive toolkit for installing and managing the NEAR Protocol CLI across all platforms.