Claude Squad: Run Many Claude Code Agents in tmux

Claude Squad is a terminal UI for running several AI coding agents at once. Each session runs in tmux inside its own git worktree, so you can start a task, let it work in the background, review the diff, and push a branch, all from one window. Claude Code is the default program; Codex, Gemini, Aider and others work too.

Who it is for

Developers who want parallel agents on independent tasks without hand-managing worktrees and terminal tabs.

Install

With Homebrew:

brew install claude-squad
ln -s "$(brew --prefix)/bin/claude-squad" "$(brew --prefix)/bin/cs"

Or with the install script, which puts cs in ~/.local/bin:

curl -fsSL https://raw.githubusercontent.com/smtg-ai/claude-squad/main/install.sh | bash

Prerequisites: tmux and gh. Run it with cs.

Workflow

• n creates a session, N creates one with a prompt.
• Enter or o attaches so you can reprompt; ctrl-q detaches.
• s commits and pushes the branch to GitHub.
• c checks out: commits changes and pauses the session; r resumes.
• D kills a session.

Use cases

• Fix three unrelated bugs in parallel and review each branch separately.
• Let one agent write tests while another refactors a different module.
• Compare Claude Code and Codex on the same task in separate worktrees.

Pros

• Isolation by default through git worktrees.
• Review-before-apply flow keeps you in control.
• Works with several agent CLIs.

Cons and caveats

• AGPL-3.0 license, which matters if you redistribute modified versions.
• The --autoyes mode is experimental and skips permission prompts; use it only in safe environments.
• Parallel agents multiply token usage and hit plan limits faster.
• Worktrees and tmux sessions linger if you do not clean them up.

Links

• Repo: smtg-ai/claude-squad
• Related: container-use for container isolation, subagents guide, permission modes