Effortlessly juggle features, bugs, and experiments, simultaneously, in parallel, all at once. Every task gets its own isolated workspace to run anything a terminal can - agents, dev servers, etc.
Built for Apple Silicon · v0.24.0
Add activity feed to groups view3m
Refactor authentication middleware1h
Fix expense model validation2hOpen this page on a wider screen to see the full three-panel layout.
Under the hood, Mooj is just a glorified terminal manager. Easily navigate through your workspaces and work on multiple tasks in parallel.
No repositories found.
Add one with the + button above.
Point Mooj at a git repo on disk. It tracks every branch, commit, and PR from there.
You get a fresh worktree at ~/mooj/workspaces/<repo>/<codename>/ to make your isolated code changes.
Open as many terminal tabs as you want, each with your full shell config and env. Run your agents, dev server, logs, etc.
Behind the UI, a workspace is just a git worktree on disk at
~/mooj/workspaces/<repo>/<codename>/. You can cd into it from any other
terminal, point your editor at it, rsync it, even delete it with
rm -rf.
Everything stays local. Mooj does not make any outbound network requests (except to check for app updates).
~/mooj/workspaces/ ├── mooj/ │ ├── shiba-inu/ (feat/activity-sidebar) │ ├── akita-inu/ (fix/login-redirect) │ └── goldendoodle/ (refactor/expense-partial) └── docs-site/ └── labradoodle/ (docs/api-coverage)
Your $PATH, your dotfiles, your env, the same shell you'd open in a stock Terminal. Run
claude or
codex in one, a dev server in the
next, psql or
tail -f in a third.
Each workspace reserves a block of ports the moment
it's created. For example, the shiba-inu workspace owns :10030-:10039,
akita-inu owns :10040-:10049, and they never overlap.

Review with Claude
Review with CodexSometimes you actually need a full fledged editor. One click opens the current workspace in your editor of choice.
Keyboard-first, vim-mode included. Every action is one chord away, and every chord is yours to remap.
New branch, new worktree, new agent — all from one keybinding.
Quickly cycle through parallel work inside of a workspace.
Mooj watches every terminal for agents running and waiting. Easily jump to the next one that needs your input.
Keybindings are configurable. Open Settings → Keybindings to change them.
Mooj runs entirely on your machine. No server, no signup, no telemetry.
No accounts
No signup, no SSO, no per-seat license. Install the .dmg and go.
Plain git on disk
Workspaces are just folders set up with git worktrees at ~/mooj/workspaces/.
No telemetry
Mooj doesn't track usage or phone home.
Your keys, your subscriptions, your tokens
Use the subscriptions you already pay for, or your own API keys.
The desktop app is Apple Silicon only for now.
Mooj doesn't phone home or upload anything. Agents you launch (claude, codex) do send to their providers — but those are your tokens and your billing relationship.
Mooj is free. AI agents cost whatever your provider charges (Claude Max, OpenAI API, etc.).
Anything you can run in a terminal. Built-in workflows use claude and codex specifically. Other agents run like normal in a terminal tab.
Yes. Mooj is free for any kind of use — personal, commercial, or otherwise.
A workspace is a worktree. Mooj adds lifecycle management, isolated port allocation, attached terminal sessions, PR + CI tracking, a UI for your dev environment.