MAO vs Claude Squad: a canvas or a terminal UI for parallel coding agents?
Claude Squad is a free, open-source terminal app that runs each coding agent in its own tmux session and git worktree. MAO is a paid desktop canvas where agents run as real terminals, hand each other tasks, test in browser and device portals and work on remote machines. Pick by where you want to watch the agents.
- Claude Squad is AGPL-3.0 open source and runs agents in tmux sessions
- Both give each agent's work its own git worktree and branch
- MAO adds agent-to-agent tasks, browser and device portals and SSH workspaces
Claude Squad is a free, open-source (AGPL-3.0) terminal app that starts each coding agent (Claude Code by default, or Codex, Gemini, Aider and others) in its own tmux session and git worktree, and lets you switch between them from one terminal window. MAO is a paid desktop app that puts each agent's own CLI on a canvas as a real terminal, with worktree layers, connections that let agents send each other tasks, browser and device portals, and remote workspaces over SSH. Choose Claude Squad if you want a keyboard-driven tool that lives inside your terminal and costs nothing; choose MAO if you want to see every agent at once and have them hand work to each other.
What is Claude Squad?
A terminal UI, installed as cs, that manages several AI coding agents in separate tmux sessions and git worktrees.
Its README describes Claude Squad as a terminal app that manages multiple Claude Code, Codex, Gemini and other local agents, including Aider, in separate workspaces, so you can work on several tasks at once. It explains how it works in three parts: tmux creates an isolated terminal session for each agent, git worktrees give each session its own branch, and a TUI lets you navigate and manage them.
From the session list you create a session with n (or N to start it with a prompt), attach to it with Enter to reprompt the agent, and detach with ctrl-q. A preview tab and a diff tab show the selected session's output and changes. s commits and pushes the session's branch to GitHub, and c commits the changes and pauses the session so you can check out the branch. The README also lists finishing tasks in the background, including an experimental auto-accept mode (cs -y).
You choose the agent with cs -p, for example cs -p "codex" or cs -p "gemini", or define named profiles in ~/.claude-squad/config.json and pick one when you create a session. It needs tmux and the GitHub CLI (gh), installs through Homebrew or an install script, and is licensed under AGPL-3.0. The latest release on GitHub is v1.0.20.
How do MAO and Claude Squad differ?
Both isolate each agent's work in a git worktree; they differ in where agents are shown and what agents can do beyond their own terminal.
When does Claude Squad fit better?
When you want a free, open-source tool that stays inside the terminal you already use.
- You want open source. Claude Squad's code is on GitHub under AGPL-3.0, so you can read it, build it and change it. MAO is not open source.
- You don't want to pay. Claude Squad has no price listed; MAO is a one-time purchase after a 7-day trial.
- You live in tmux and the keyboard. Each agent is a tmux session and every action is a key in one terminal window, with nothing else to open.
- You want agents to keep going unattended. The README lists finishing tasks in the background, with an experimental
-yflag that accepts prompts automatically for Claude Code and Aider. - You push straight to GitHub.
scommits and pushes a session's branch from the list, using theghCLI it already requires. - You run Aider or another agent with a custom command.
-pand profiles launch any program, such asaider --model ollama_chat/gemma3:1b. In MAO, Aider would run as a custom command, since it is not one of the listed providers.
When does MAO fit better?
When you want to see all the agents at once and have them work together, test what they build, or run on another machine.
- Every agent is visible. Each agent is its own terminal node on a canvas, so you watch several sessions side by side, zoom out to status cards, and jump to the next agent that needs attention.
- Agents hand work to each other. Draw a connection and an agent can send a task to its peer with
mao ask reviewer "..."and get the result back, instead of you copying between sessions. - Agents can check their own work. A browser portal puts a Chromium page on the canvas that a connected agent can click, type into, read and screenshot; device portals do the same for an Android emulator or phone, and for the iOS Simulator on macOS.
- The project lives on a server. A remote workspace keeps the repository, terminals and agents on a Linux machine over SSH, or inside WSL, while the canvas stays on your desktop.
- Work lands with checks. Landing a layer checks for conflicts and uncommitted changes and shows the destination branch before it merges or fast-forwards, and never resets, forces or stashes.
- Each agent keeps its own account and role. Saved provider accounts, rotating API keys and reusable Markdown roles are set per agent node.
How do I compare them on a real task?
Give both the same two-part task and see which one leaves less for you to do by hand.
- Pick a change with two independent parts, such as an API endpoint and its tests, in a repository you can throw away.
- In Claude Squad, run
cs, pressntwice to create two sessions, prompt each agent, and review each one's diff tab before pressingsorc. - In MAO (the 7-day trial is enough), create a layer for the change, place two agents on it, connect them, and have one ask the other to review its work with
mao ask. - Land the layer into Ground and compare: how many times you switched windows, copied text between agents, and checked the result yourself.
Frequently asked questions
Yes, for running several coding agents in parallel on isolated git worktrees. Claude Squad does it inside one terminal with tmux; MAO does it on a desktop canvas and adds agent-to-agent tasks, browser and device portals and remote workspaces.
Yes. Claude Squad's repository at github.com/smtg-ai/claude-squad is licensed under the GNU AGPL-3.0, and its README lists no paid plan.
Yes. Its README launches Codex with cs -p "codex" and Gemini with cs -p "gemini", and profiles in the config file let you pick an agent per session. Claude Code is the default program.
Yes. tmux and the GitHub CLI (gh) are listed prerequisites, and each agent runs in its own tmux session.
Yes. Claude Squad gives each session a git worktree on its own branch. In MAO, each layer is a Git worktree on its own branch with its own canvas, and several agents can work on one layer.
Sources
- Claude Squad README: supported agents, tmux and git worktree isolation, the key commands,
-p,-yand profiles, prerequisites, installation and the AGPL-3.0 licence. - Claude Squad website: supervising several agents in one UI, isolating tasks in git workspaces, reviewing work before shipping.
- Claude Squad licence: GNU Affero General Public License v3.
- Claude Squad releases: v1.0.20 with builds for macOS, Linux and Windows on amd64 and arm64.
- MAO product facts: platforms, supported agents, price and free trial.
- Run a shell or agent and Use a structured ACP chat: terminal nodes, presets and custom commands, the optional chat view.
- Create a worktree layer and Land a layer into Ground: layers as Git worktrees and the checks before landing.
- Send a task and collect its result:
mao ask,mao send,mao nextandmao resultbetween connected agents. - Browser and device portals and View an iOS Simulator portal: browser, Android and iOS Simulator portals.
- Remote workspaces: Linux servers over SSH, WSL, and that remote workspaces come with the licence.
- Add and choose provider accounts, Rotate an account's API keys and Save and assign agent roles: per-agent accounts, key pools and roles.
Checked on 2026-10-28.
Questions? Ask in the Discord or write to hello@miguer.com.