More
Asked and answered.
Everything people ask about the fleet, the Conductor, keys, limits, and privacy. When an answer deserves a full page, the link is right there.
The machine
The fleet and the Conductor.
What exactly is the fleet?
Six real Claude Code CLI sessions running in parallel, signed in with your existing Claude login. They are real terminals, not a simulation: you can watch any of them or type into one whenever you like. The fleet page goes deeper.
What does the Conductor do all day?
It plans your goal into a Mission DAG of tasks, dispatches them across the fleet, watches every terminal, answers permission prompts, recovers stalls and rate limits, and verifies work before it counts as done.
Does the fleet need an API key?
No. The fleet runs entirely on your Claude login. Only the Conductor takes a key, and even that is optional if you run it locally. See bring your own key.
Which models can be the Conductor?
OpenAI, Anthropic, Google, xAI, Groq, Cerebras, DeepSeek, Mistral, OpenRouter, any OpenAI-compatible endpoint, or a local model through LM Studio, MLX, or Ollama.
Safety
Control, limits, and privacy.
What does auto-accept actually press?
Only the safe affirmative: the conservative yes for the specific action in front of it, never a broader grant. It is toggleable, and shift+tab cycles one terminal's permission mode. Full story on the auto-accept page.
What happens when a session hits a usage limit?
Loom parks that Claude account and rotates to the next pooled one, then recovers the session and re-sends the task. See account rotation and the rate limits guide.
Where do my keys live?
In your operating system's keychain, nowhere else. Outbound calls go through an SSRF-guarded Rust proxy, and a secret-path deny-list guards reads and writes. The security page has the architecture.
Does Loom phone home?
No. There is no telemetry and no account beyond your Claude login. The privacy page says the same thing in legal language.
Practical
Platforms, price, and the alternatives.
Which platforms does it run on?
macOS, Linux, and Windows, with WSL available per terminal tab on Windows. The installer is a Tauri 2 app of about 13 MB.
How is this different from plain Claude Code?
Claude Code is excellent on its own, and Loom is built on exactly that. Loom runs six of them at once and adds a planner that dispatches, supervises, and verifies, plus a full workspace around the terminals. The honest comparison is at Loom vs plain Claude Code.
How does it compare to Cursor, Devin, or the rest?
Each tool has real strengths and we wrote those up fairly. Start with alternatives to Loom and follow the comparison that matches what you use today.
What does it cost?
The app is free and Apache-2.0, forever. You pay your model providers directly, and a local Conductor costs nothing. The pricing page itemizes it.
Hand it the work.
Walk away.
macOS, Linux, and Windows. Around 13 MB. Free and open source.