Loom v0.7.4 is out for macOS, Linux and Windows

Guides

Zero to fleet in about a minute.

Install Loom, sign into claude once, pick a Conductor, type a goal. That is the entire setup.

1. Install Loom

Grab the build for macOS, Linux, or Windows from GitHub releases. The installer is around 13 MB, free, and Apache-2.0. Platform details live on the install page.

2. Sign into claude once

Loom conducts six real Claude Code CLI sessions, so it needs the official CLI on your machine. Install it, run claude, and log in. Every session in the fleet reuses that login. There is no API key for the fleet, and no Loom account either. The Claude login guide explains why.

3. Pick a Conductor

The Conductor is the reasoning model that plans your goal into tasks and runs the mission. Bring your own key from OpenAI, Anthropic, Google, xAI, Groq, Cerebras, DeepSeek, Mistral, OpenRouter, or any OpenAI-compatible endpoint, or run it fully local on LM Studio, MLX, or Ollama. Keys are stored in the OS keychain only. See Bring your own key.

4. Type your first goal

Describe an outcome, not a procedure. The Conductor plans it into a mission DAG, dispatches tasks across the six terminals, and verifies the work before it counts. The first goal guide shows what good goals look like.

The only terminal step

Two commands, once.

This is the whole command-line portion of setup. On Windows there is a PowerShell equivalent, covered on the Windows page.

shell
# install the official Claude Code CLI
$ curl -fsSL https://claude.ai/install.sh | bash
# sign in; the whole fleet reuses this login
$ claude

Then what

You watch. It works.

It answers prompts

Claude Code asks before edits and commands. Auto-accept presses only the safe affirmative so six sessions never sit waiting. It is toggleable, and shift+tab cycles one terminal's permission mode.

It recovers

Rate limits and stalls are detected and handled, and pooled Claude accounts rotate when one hits a usage limit. The mission keeps moving without you.

It verifies

Finished is a claim until it is checked. The Conductor re-checks work before a task counts as done, and the mission DAG and live activity strips show exactly where things stand.

Hand it the work.
Walk away.

macOS, Linux, and Windows. Around 13 MB. Free and open source.