A terminal that runs your coding agents while you're away.
Nightcrew hands each task you write to Claude Code, Codex or another agent from the ACP registry in its own git branch, checks the work before calling it done, and leaves you a short report in one inbox.
v0.2.0 · Apple silicon · macOS 13 or later · 71 MB
Screens from the app running on example projects (a shop's web front end, its API and its docs).
Every task gets its own branch
Nightcrew makes a git worktree and branch for each task, so several agents can work at once without touching each other's files or yours. In a background run the agent works inside that worktree without asking before every edit.
Background runs refuse force-push, mirror push and deleting remote branches.
Push and merge follow the rules you set for the project.
A folder that holds several repos (front end and back end) works too.
new task · its own worktree and branch
It checks the work before it says done
Changes run through harness, a set of skills, checks and reviewer agents shipped with the app. A run writes a short spec, works in small steps with a test and a commit each, and sends the result to reviewers before it reports.
When a step needs a decision only you can make, the run stops and says so on its card, instead of guessing.
the engine is done; the run waits for your answer: backfill old orders or start empty?
Use the agents you already have
Claude Code and Codex are built in. Settings → Engines lists the official ACP registry, so you can add another agent in one click; the bot picks one per task by its size and the quota left on your accounts.
45 in the registry today, among them Gemini CLI, Grok Build, GitHub Copilot, Cursor, OpenCode, goose, Qwen Code, Kimi Code and Mistral Vibe.
settings · engines from the ACP registry
Before you install
macOS 13 or later on Apple silicon. Linux is next, Windows later.
Not notarized yet. macOS stops the first open; the download section says how to allow it.
You need a Claude account. The bot runs on Claude Code; agents use your own subscriptions.
Checked runs need Claude Code. Other agents do the work too, with an approval card for each step.
Some agents ask for their own login the first time, in their own command line.
Your code stays on your Mac. Agents send what they read to their own providers, as they do outside Nightcrew.
Changelog
0.2.0 · 12 November 2026
Settings → Engines: install agents from the ACP registry.
Projects can be a plain folder holding several repos.
Code changes run as background autopilot runs, with progress on the card.
One inbox across projects; the title bar counts what needs you.
harness ships inside the app; adding a repo offers to register it.
The bot's context stays small and starts over from a short brief.
Renamed from Fleetty to Nightcrew; first public disk image.
0.1 · 11 November 2026
Terminal with command blocks, a bot that delegates to Claude Code and Codex, one worktree per session.
This is a beta: it works for daily use, but expect rough edges and changes between builds. macOS 13 or later on Apple silicon. The disk image includes Node.js, an editor and the harness kit. Linux and Windows packages come next.
First open. This build is not notarized yet, so macOS stops it the first time. Drag Nightcrew to Applications, open it once, then go to System Settings → Privacy & Security and click Open Anyway. Or run:
A bot runs on a Claude model you pick, in Claude Code print mode with read-only built-in tools (read, search, fetch) and only the MCP servers Nightcrew passes it. The engines below it can be any of the agents you installed. Its rules for each message:
Any file changeDelegated as an autopilot run. The bot writes nothing before delegating.
A questionAnswered from the code, without starting a run.
A skill command/something is passed to the engine unchanged, so your own skills keep working.
Engine, model, accountChosen by task size and quota left; the reason is shown on the card.
The answerEnds with Done, Left and Needs you lines, which feed the status strip and the inbox.
# what an engine receives for "Add CSV export to the orders page"/autopilot Add CSV export to the orders page: same filters as the table, at most 50,000 rows.
Isolate
One worktree per task
Each session works in <repo>/.worktrees/<branch> on its own branch from the base you chose. .env files are copied in and the project's setup script runs once. Two autopilot runs never share a worktree, so parallel agents cannot overwrite each other or your own checkout.
main→feat/orders-csv→.worktrees/feat-orders-csv→3 commits, one per slice
Run
Autopilot: spec, slices, checks, review
Autopilot is a harness skill. Harness is a folder of agent context shared by all your projects: skills, reviewer agents, hooks, standards per stack, and a record per run. A run goes through these stages, and the card follows them by reading the run record every 15 seconds:
SpecWhy, scope, one requirement per slice with given / when / then scenarios.
Slice by sliceFailing test, the smallest change, the fast check, the outcome observed for real, one commit.
ReviewDiff, security and edge-case reviewers sized to the change, plus a browser walkthrough for UI work.
StopDone when every slice passes and the full check is green; blocked when only you can decide.
Autopilot needs Claude Code through the ACP adapter and a repo registered in harness. Other engines and unregistered repos still run tasks, with normal approval cards.
Guard
Bypass, with fences
A background run uses bypass permissions so it never waits for you. What it can do is still limited:
Refused in background runsgit push --force, -f, +ref, --mirror, --delete, :ref and --prune pushes in background runs, whatever the engine. In an ordinary session they come to you as an approval card.
Push and mergeOnly when your project's harness rules allow it, through the run's own ship step. Otherwise everything stays local.
Ask beforeRules in the bot profile still go to you as approval cards.
SecretsAccount sign-ins and MCP tokens live in the macOS Keychain, never in the database, logs or command lines.
Report
The inbox and the notification
When an engine ends, the bot reads its report and the run's outcome, writes Done, Left and Needs you, and sends a macOS notification: done with the slice count, or blocked with the reason. The inbox regroups every few seconds.
Needs youa Needs you line, an approval waiting, or a blocked run
Failedthe session or its engine failed
Runningan engine is working
Donethe rest; read once opened
Context
Context that stays small
The header shows the bot's context, measured from the model's own usage, against a limit: 200k tokens, or 60% of a smaller model window. It turns amber at 70%. When the conversation reaches the limit, and has grown at least a quarter of the limit past its fresh size, the next turn starts a fresh conversation from a brief.
fresh · 94kamber · 140klimit · 200k → fresh
Session brief:
- Task: Add CSV export to the orders page
- Branch: feat/orders-csv (base main)
- Done: CSV export on /orders, 3 slices
- Needs you: review the branch and open the PR
- Latest engine run: Claude Code · sonnet (done) · autopilot · 3/3 slices · done+ the last 20 messages
Long runs
Long runs
An engine prompt may run for up to 12 hours. If it still times out, the main run continues once on its own. Engines can also consult an advisor model (Fable 5.2 by default, set in Settings → Advisor) when they have a concern, instead of stopping to ask you.
Docs / Setup
Install and set up
About ten minutes on a Mac that already has git. Steps one to five happen once per machine, step six once per repo.
Open the disk image and drag the app to Applications. Builds are not notarized yet, so macOS stops the first open: go to System Settings → Privacy & Security and click Open Anyway. When macOS asks whether the app may use the Keychain, allow it: sign-ins and tokens are stored there.
The disk image already contains Node.js, a terminal editor, the notification helper and the harness kit, so a bare Mac needs only git and Python 3.11+ on top.
2
Let Setup check the machine
The first launch opens Settings → Setup. Every tool shows as found (with its path or version) or missing, with an Install or Sign in button that runs in a terminal you can watch. Re-check runs it again; when PATH changed, a Restart bar appears.
settings · Setup · here the Claude sign-in is still missing
Tool
Needed
What it is for
git
required
Worktrees and branches. Xcode's command-line tools are enough.
Claude Code CLI
required
The bot runs on it, and it is the default engine.
Claude sign-in
required
An account for the bot and engines. Sign in opens Accounts.
Node.js / npx
required
Runs the ACP adapters. Bundled in the app.
Claude ACP adapter
optional
@agentclientprotocol/claude-agent-acp, fetched by npx on first use. Needed for autopilot.
GitHub CLI (gh)
optional
Pull requests from Source Control.
Docker
optional
Projects whose checks start databases or services.
Python 3.11+
optional
Only needed to install harness. macOS ships 3.9, which is too old.
harness
required
Install copies the bundled kit, links its skills, agents and hooks into ~/.claude and puts harness on PATH.
3
Sign in your accounts
Settings → Accounts lists Claude and Codex accounts. The default one is your normal Claude Code login; add more to spread work across quotas. Each account keeps its own config folder, and the bot runs a task on the account with the most quota left.
/loginOpens Accounts on the account that needs to sign in, from any session.
/accounts add claudeConnect another Claude account. /accounts add codex for Codex.
/accounts remove <id>Remove an added account. The default login stays.
/accounts refreshRead every account's quota again.
4
Connect engines
Claude Code (@agentclientprotocol/claude-agent-acp) and Codex (@zed-industries/codex-acp) are built in and start through npx. Any agent from the ACP registry installs from Settings → Engines in one click; an agent that is not listed is one block in the config file; it then appears in the bot profile's engine list, and the bot can pick it.
Then choose, per bot, which engines it may use (Settings → Bots). The permission mode for a session cycles with shift+tab in the message box:
Ask permissionsEvery edit and command becomes an approval card.
Accept editsFile edits go through, commands still ask.
Plan modeThe engine plans and does not change files.
Bypass permissionsNo prompts, inside the session's worktree. Used by background runs.
The mode is remembered per session, per project or for everything. When something needs approval, a card offers Approve, Always for this bot (saved as a rule) or Deny, and a macOS notification tells you it is waiting.
5
Give agents your tools (optional)
Settings → MCP servers: pick GitHub, Sentry, Linear, Notion, GitLab, Playwright or Filesystem, or a custom HTTP server or command. Paste its token (Linear and Notion sign in through the browser) and choose whether this project or all projects get it. The token goes to the Keychain; sessions get the server from their next turn.
settings · MCP servers · scope "This project"
6
Add a repo and register it with harness
Add a project folder with the + in the sidebar. If harness does not know it yet, you are asked once, with a key from the git remote (owner/name, or local/name without a remote) and layers guessed from the files it finds: Next.js, React, NestJS, .NET, Kotlin Spring, Postgres, Kubernetes. Register, or Not now. Setup lists every project and lets you register later.
register · prefilled from the repo
What registering gives you. Engines in this repo get the project's context, checks and rules, and every code change runs as an autopilot run with a spec, slices and reviewers. Without it, tasks still run, with approval cards.
7
Give it tonight's task
Start a session on the repo (the + next to the project) and write the task. Then close the window or the laptop. A notification arrives when the run is done or blocked; the inbox has the rest.
you Add CSV export to the orders page. Same filters as the table, max 50k rows.
bot Delegated to Claude Code · sonnet as a background autopilot run
worktree feat/orders-csv · 0/3 slices · starting… 47 minutes later …Done: CSV export on /orders, 3 slices, checks green, walked in a browser
Left: nothing
Needs you: review the branch in Source Control and open the PR
Handy in a session
Enter · Shift+EnterSend the message, or start a new line.
/ then TabComplete a command, including your own Claude Code skills and plugins.
⌘V · drop filesAttach a screenshot or files to the message.
remember that …Saves a line to the bot's memory for every later task.
terminal ↗On an engine card: a shell in that worktree. Type # question there to ask the bot about the last command.
⌘B · ⌘⌥BHide or show the sidebar and the right panel.
nightcrew 
Download