What Claude Squad does well, and where it stops
Claude Squad's latest release is v1.0.20, from August 20, 2026, and it's still maintained. It runs Claude Code by default and other terminal agents with -p, such as Codex or Aider.
- Good: one tmux session and one git worktree per task, so agents never edit the same files.
- Good: review each instance's diff and check out its branch before you push.
- Good: an experimental
-y(auto-yes) mode accepts prompts so tasks can finish in the background. - Limit: it needs tmux and the GitHub CLI; the README doesn't mention Windows, where tmux means using WSL.
- Limit: it's a terminal UI. There's no board, no shared view for a team, and you start every task yourself.
How to run several agents in parallel, with or without Claude Squad
Give every task its own worktree
That's the core of Claude Squad, and Claude Code now does it itself:
claude --worktree <name>starts a session on its own branch and checkout. See the Claude Code worktree guide.# Claude Squad: Claude Code by default, or another agent with -p cs cs -p "codex" # Claude Code on its own: one worktree and branch per session claude --worktree fix-login claude --worktree add-billingPick how you want to watch them
A terminal list (Claude Squad), a desktop app (Paradev, Conductor while it lasts) or a kanban board of sessions (Kanban Code). See Claude Code kanban boards.
Decide who starts the next task
In all of these you start each session. If you want agents to take the next issue themselves, give them a schedule or a heartbeat; see how to make Claude Code proactive.
Review before you merge
Whatever runs the agents, keep merging with a person: read each diff, run the checks, then merge.
Claude Squad alternatives compared
Facts from each project's own site or repository, checked in November 2026.
| Interface | Agents | Platforms | Who starts tasks | License | |
|---|---|---|---|---|---|
| Claude Squad | Terminal UI (tmux) | Claude Code, Codex, Gemini, Aider | Needs tmux and gh; Windows via WSL | You | AGPL-3.0 |
| Claude Code on its own | claude --worktree, parallel sessions in the desktop app | Claude Code | macOS, Windows, Linux | You (or a schedule) | Commercial (Anthropic) |
| Paradev (formerly Superset) | Desktop app, a worktree per task; automations | Claude, Codex, Cursor Agent, OpenCode and any CLI agent | macOS; Linux experimental; no Windows yet | You, or an automation | MIT |
| Kanban Code | Kanban board of Claude Code sessions | Claude Code | macOS 26 app, Windows app | You | Apache 2.0 |
| Conductor | Mac app for parallel agents; joining Meta, cloud ends November 20, 2026 | Claude Code, Codex, Cursor, OpenCode | macOS | You | Commercial |
| Crewmark | Hosted board with AI employees that wake on a heartbeat | Claude Code, Codex, DeepSeek Harness, OpenCode | Web board; daemon on macOS, Windows, Linux | The employee takes the next issue | Commercial |
Claude Squad vs Conductor
Both run several agents in parallel worktrees. Claude Squad is a free, open-source terminal app; Conductor was a polished Mac app with a review flow and cloud workspaces. Conductor is joining Meta: new sign-ups are closed, its cloud shuts down on November 20, 2026 and the local app gets updates only until early 2027. If you were choosing between the two today, Claude Squad is the one still being developed.
Which Claude Squad alternative fits you
- You only use Claude Code: try its built-in
--worktreeflag and the desktop app's parallel sessions before adding a tool. - You want a desktop app for many agents on a Mac: Paradev.
- You want to see sessions as cards: Kanban Code, or read Claude Code kanban boards.
- You're on Windows: Claude Code's CLI and desktop app run natively; Claude Squad needs WSL.
- You want agents to keep working without you starting each task, with a board your team shares: Crewmark (below). Also see multiple Claude Code agents.
From a squad you drive to a crew that works on its own
Claude Squad makes parallel agents easy to run, but each one waits for you to start it. Crewmark turns a Claude Code, Codex, DeepSeek Harness or OpenCode session into an AI employee with a role, a goal and working hours. It wakes on a heartbeat, takes the next issue from a shared board, works in its own checkout and messages you when it needs a decision, on the subscription you already pay for.
Connect your repository and agent
Run the setup command in the project. Several employees can share a repository, each in its own checkout.
Give each one a role
An engineer, a reviewer, an SEO writer, each with a goal. See AI employee roles.
Fill the board and review
Every 15, 30 or 60 minutes during working hours each employee takes the next issue; you review pull requests and answer questions.
Claude Squad FAQ
What is Claude Squad?
An open-source terminal app (AGPL-3.0) from smtg-ai that runs several coding agents at once, each in its own tmux session and git worktree, so you can review each change before pushing.
Does Claude Squad work with Codex?
Yes. Run it with cs -p "codex" after setting your OpenAI API key. It also launches Gemini and Aider, and Claude Code by default.
Does Claude Squad work on Windows?
Its README doesn't mention Windows and it needs tmux, so on Windows you'd run it in WSL. Claude Code's own --worktree flag and desktop app run on Windows natively.
Is Claude Squad still maintained?
Yes. Its latest release is v1.0.20, from August 20, 2026.
Claude Squad or Conductor?
Conductor is joining Meta and winding down, with its cloud ending on November 20, 2026. Claude Squad is open source and still developed.
Do I still need Claude Squad now that Claude Code has worktrees?
Not for isolation alone: claude --worktree gives each session its own branch. Claude Squad still helps if you want one terminal view to manage many sessions and agents other than Claude Code.
Checked against the Claude Squad README and releases and each alternative's own site or repository on November 7, 2026.