A terminal multiplex for AI agents that refuse to multitask

What it does

Claude Squad is a Go-based terminal interface that spawns and manages multiple AI agent sessions—such as Claude Code, OpenAI Codex, Gemini CLI, or Aider—inside isolated environments. It wraps each agent in its own tmux session and git worktree, giving every task a separate branch and working directory. You can kick off background jobs, pause and resume sessions, review diffs, and push commits without leaving the TUI.

The interesting bit

The project is essentially well-crafted glue code around tmux and git worktrees, which is exactly why it works: instead of reinventing isolation, it borrows it from standard Unix tools. This lets you treat agents as cattle rather than pets, running an experimental auto-accept mode so they can grind through tasks unattended while you review the diffs later.

Key highlights

• Supports multiple agent backends—Claude, Codex, Gemini, Aider, or custom shell commands—via configurable profiles.
• Each session lives in an isolated git worktree, eliminating branch conflicts between concurrent tasks.
• Built-in diff preview and commit workflow: review changes, check out (pause), resume, or push to GitHub from the interface.
• Background execution with an experimental --autoyes flag for unattended agent runs.
• Requires external dependencies: tmux and the GitHub CLI (gh).

Caveats

• Session startup can fail with a tmux timeout if the underlying agent binary is outdated, according to the FAQ.
• The auto-accept feature is explicitly marked experimental.
• It is fundamentally an orchestration layer; if you do not already use tmux or agent CLIs, the tool stack adds friction.

Verdict

Grab it if you are already using Claude Code or Codex and need to parallelize tasks across features or codebases. Skip it if you prefer a single agent in a single directory, or if wrangling tmux sessions and GitHub CLI dependencies sounds like more overhead than help.