Skip to content
← Tools / Skills
Toolsmtg-ai/10 trustMetrics-checked

Claude Squad

Run a squad of Claude Code, Codex, and Aider agents in parallel — each in its own tmux session and git worktree.

Visit official site →

About

Claude Squad is a terminal app (Go, Bubble Tea TUI, cs binary) for running multiple AI coding agents at once without them stepping on each other. Press n, pick a program — Claude Code, Codex, Gemini CLI, Aider, OpenCode, Amp, or any custom command — describe the task, and each agent gets an isolated tmux session plus its own git worktree, so parallel tasks never conflict.

You watch every agent's progress in one window, flip between live preview and git diff tabs, review changes before applying them, and commit or push per worktree when a task lands. An experimental auto-yes mode lets trusted runs accept prompts unattended, and a Next.js web interface ships alongside the TUI. With 8,600+ stars it is one of the more popular agent-orchestration TUIs, though the AGPL-3.0 license deserves a read if you plan to modify or network-deploy it.

What it can do

  • Parallel agents (Claude Code, Codex, Aider, Gemini, OpenCode, Amp, custom commands)
  • Isolated tmux session and git worktree per agent — no cross-task conflicts
  • Real-time TUI monitoring with preview/diff tabs and change review before apply
  • Experimental auto-yes mode for unattended runs
  • Commit, push, checkout, pause, and resume per session
  • Companion Next.js web interface

Setup

  1. Install tmux and the gh CLI (prerequisites).
  2. Install via brew install smtg-ai/claude-squad/claude-squad, go install github.com/smtg-ai/claude-squad@latest, or the install script.
  3. Run cs to open the TUI.
  4. Press n to spawn an agent, Enter to inspect it, Tab to switch preview/diff, s to commit and push.

Prerequisites

tmux and the GitHub `gh` CLI; at least one supported agent installed (Claude Code, Codex, Aider, Gemini CLI, etc.).

Use cases

coding-agentsterminaldeveloper-toolsopen-source

Works with

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: 8,608 GitHub stars (GitHub API), not archived, pushed Aug 20 2026, feature set corroborated across the canonical repo and multiple forks. Strengths: real adoption, clean isolation model (tmux + worktrees), change review before apply. Docked for: (1) AGPL-3.0 — fine for personal use, but check with legal before modifying or offering it over a network; (2) last push ~2 months ago, slower cadence than peers; (3) auto-yes mode is genuinely risky — unattended agents with shell access need tight scoping. License: AGPL-3.0.

★ 8,608 GitHub stars

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
7.0/10
Maintenance · recency, releases, issue load
7.0/10
Openness · license + docs
9.0/10
Manual review · sanity read + red-flag check
7.0/10
  • · 8,608 GitHub stars (log-scaled vs catalog peers, verified via GitHub API); not archived
  • · last push Aug 20 2026 — slower cadence than peers; AGPL-3.0 (OSI-approved, network-copyleft — check with legal before modifying or network-deploying)
  • · tmux sessions + git worktrees per agent, TUI with preview/diff review before apply — feature set corroborated across canonical repo and forks
  • · manual review: experimental auto-yes mode is genuinely risky — unattended agents with shell access need tight scoping

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.