The board
Every worker, live — what it's doing, how far along, and whether it's stuck. A control room, not a terminal.
Runs on your Claude Max plan — no API meter
You bring the vision. The fleet builds it.
Claude, orchestrating Claude.
Fleet is a macOS app that runs a fleet of Claude Code agents under one captain AI — it writes the prompts, runs the workers, verifies the design and the code, cleans up after itself, and surfaces only what needs you.
Apple Silicon · macOS
One captain. A whole fleet of AI agents. Running on your flat Claude Max plan — never a metered API bill.
AI that prompts AI
You stopped writing the code. Now you can stop writing the prompts.
Fleet's captain reads your repos, researches what it needs, and writes the ideal brief for every task — then hands each one to its own AI worker. You describe the outcome; it figures out the instructions.
The whole loop — autonomous
Describe it once. It runs to done.
01
You set the vision
Say what to build in plain English — a feature, a fix, a whole site. No tickets, no prompt-engineering.
02
The captain runs the fleet
It writes a brief per task and spawns an AI worker for each — in its own terminal, across all your projects, in parallel.
03
Only what matters reaches you
It drives each worker to a high bar, verifies the work, cleans up when they finish — and pings you only for the key approvals and decisions.
One captain, a whole fleet
See everything. Touch almost nothing.
Full visibility into every worker — and a calm inbox for the handful of moments that genuinely need you.
Approval inbox
Risky or out-of-scope calls pause and surface as a card — approve or redirect from your desk or your phone.
Worktree isolation
Parallel workers each get their own branch, so a dozen agents never collide on the same files.
Schedules
"Every morning, sweep for bugs." Recurring jobs run on their own and deliver the result — even while you're away.
Skills
Save a worker that nailed it as a reusable recipe — and re-run that proven playbook on demand.
Cleans up after itself
Finished workers are verified, parked, and swept off the board automatically. The fleet stays tidy without you.
The difference
An orchestrator with taste.
Most agents can write code that compiles. Fleet judges whether the result is actually good — the design as much as the logic — before it ever reaches you.
01A separate critic model grades every design against an anti-slop bar — the judge is never the generator.
02Work isn't "done" until it's proven — a real check passes, a screenshot looks right, the endpoint responds.
03Great code and great design, with the only human input being "what should we build."
The economics
A 24/7 fleet that doesn't bill by the token.
Fleet drives your interactive Claude sessions — so a fleet running all day stays on your flat Max subscription instead of a metered API meter that punishes the agents for working.
Fleet
Flat Claude Max plan. Run the fleet all day — the cost doesn't move.
Metered cloud agents
Per-token billing. The more they build, the more it costs.
Questions & answers
What to know before you download.
Updated June 2026
What is Fleet?
Fleet is a macOS desktop app for running a fleet of Claude Code AI agents. One captain AI orchestrates many worker agents — it writes each prompt, runs the workers in parallel, verifies the code and the design, and surfaces only the decisions that need you. (It's a desktop app for Claude Code agents — not FleetDM or JetBrains Fleet.)
How does Fleet work?
You describe an outcome in plain English, and Fleet's captain AI does the rest. It reads your repositories, writes a tailored brief for each task, and spawns a separate Claude Code worker for every job — each in its own terminal and its own git worktree, running in parallel across your projects. The captain drives each worker to a high bar, verifies the result, cleans up, and pings you only for the key approvals.
How is Fleet different from metered cloud AI agents?
Fleet runs on your flat Claude Max subscription instead of a metered API. It drives your interactive Claude sessions locally on your Mac, so a whole fleet working all day costs the same as a single session. Metered cloud agents bill per token — the more they build, the more they cost.
What does Fleet cost?
Fleet is a free download today and runs on the Claude Max plan you already pay for, so there's no per-token API bill. You bring your own Claude Max subscription; Fleet adds no metered usage cost on top of it.
Do I need an API key to use Fleet?
No. Fleet drives the interactive Claude sessions on your existing Claude Max plan, so there's no API key to create or manage and no metered API billing to set up.
What can non-developers build with Fleet?
Fleet is built for non-developers. You describe what you want — a feature, a fix, or a whole website — in plain English, and the captain AI handles the prompts, the code, and the verification. Because you set the vision instead of writing the instructions, you can ship software without prompt-engineering or writing code yourself.
What platform does Fleet run on?
Fleet is a native macOS desktop app for Apple Silicon Macs. You download it and run your agent fleet locally on your Mac.