tmux-IDE: A Terminal-Based Multi-Agent IDE for Claude
What tmux-IDE Does
tmux-IDE is a lightweight, terminal-based IDE that prepares Claude agent-team layouts in one terminal session. It builds a lead pane, teammate-ready Claude panes, and development tools in a single tmux layout. The tool enables the environment, then Claude forms the team after you prompt it.
Key Features and Setup
The tool is version 1.3.0 and can be installed multiple ways:
$ curl -fsSL https://tmux.thijsverreck.com/install.sh | shnpm i -g tmux-idenpx tmux-idefor instant trial
The install script registers the Claude Code skill automatically, allowing Claude to configure your workspace automatically.
How It Works
tmux-IDE prepares panes and enables agent-team mode. From there, you prompt the lead Claude agent to organize the team and assign work in natural language. The lead coordinates the team while teammates work independently in their own panes with focused tasks.
Agents communicate through shared tasks and messages. The lead assigns tasks, teammates claim them and report back. Once the layout is running, the lead can recruit teammates, reassign work, and reshape the workflow through normal Claude prompting.
Configuration and Usage
Layouts are defined declaratively in YAML:
ide.yml
name: my-project
rows:
- size: 70%
panes:
- title: Claude 1
command: claude
- title: Claude 2
command: claude
- panes:
- title: Dev Server
- title: ShellQuick start commands:
$ tmux-ide init --template agent-team→ Creates ide.yml with agent team layout$ tmux-ide→ Launches IDE session with lead and teammate-ready panes$ tmux-ide restart→ Restarts with updated layout
The tool auto-detects stacks including Next.js, Vite, Python, and Go, running dev servers alongside your agent team. Layouts are reproducible across machines and projects.
This tool is for developers who want to create always-on coding setups using SSH and tmux, particularly useful for long-running coding tasks where you want to start work, give instructions to AI agents, and disconnect while they continue working.
📖 Read the full source: HN AI Agents
👀 See Also
New Tool Injects Instructions into Claude Code Based on Context Usage
A developer created a tool that monitors context usage and injects custom instructions to prevent mid-task compaction disasters.
Code Evolution Method Triples LLM Performance on ARC-AGI-2 Benchmark
Researchers achieved a 2.8x improvement on the ARC-AGI-2 benchmark using code evolution with open-weight models, reaching 34% accuracy at $2.67 per task. The same method pushed Gemini 3.1 Pro to 95% accuracy at $8.71 per task.
OpenClaw Smart Router Open-Sourced for Automatic Model Selection
A developer has open-sourced a Smart Router for OpenClaw that automatically classifies queries by complexity and routes them to optimal models, saving 60-80% on API costs compared to always using premium models like Claude or GPT-4o.
nah: A context-aware permission guard for Claude Code
nah is a PreToolUse hook that intercepts every tool call in Claude Code, classifying commands by action type like filesystem_read or git_history_rewrite and applying policies based on context. It runs a deterministic classifier in milliseconds with optional LLM escalation for ambiguous cases.