Resources
The fleet, defined.
Twelve terms you will meet in the app, the guides, and the README, each explained in plain English.
Fleet
The six real Claude Code CLI sessions Loom runs in parallel. Not a simulation or a wrapper: each one is the genuine CLI in its own native PTY, signed in with your existing Claude login, no API key required. More on the fleet page.
Conductor
The reasoning model that sits above the fleet. It plans a goal into tasks, dispatches them across the six sessions, watches every terminal, answers permission prompts, recovers rate limits and stalls, and verifies work before it counts. You bring its key. Details on the Conductor page.
Mission
One goal, end to end: from the brief you write to the verified finish. A mission's progress is visible as a mission DAG and live activity strips.
Slot
One of the six positions in the fleet. Each slot holds a single Claude Code session that the Conductor can hand tasks to, and each has its own terminal you can open at any time.
Brief
The goal in your own words. The Conductor turns the brief into tasks, so the clearer it states what done looks like, the better verification can hold the fleet to it. The first goal guide walks through writing one.
Workspace
A folder you have explicitly authorized Loom to work in. Every file, git, and shell operation is checked against the workspace authorization registry; outside it, nothing runs. See workspaces.
Auto-accept
The toggleable behavior that answers Claude Code permission prompts for you. It presses only the safe affirmative, never anything broader, and shift+tab cycles one terminal's permission mode when you want manual control of a single session. See auto-accept.
Rotation
What happens when a pooled Claude account hits its usage limit: Loom rotates to another account in the pool so the mission keeps moving instead of waiting for the limit to reset. See account rotation.
Verification
The Conductor's check that a task actually did what it claimed before it counts toward the mission. Work that fails the check does not silently pass. See verification.
BYOK
Bring your own key. The Conductor runs on a key you supply, from OpenAI, Anthropic, Google, xAI, Groq, Cerebras, DeepSeek, Mistral, OpenRouter, any OpenAI-compatible endpoint, or a local LM Studio, MLX, or Ollama model. Keys are stored in the OS keychain only. The fleet itself needs no key at all, just your Claude login. See BYOK.
Activity strip
The live strip that shows what each session is doing right now, so one glance across the fleet tells you whether anything needs your attention.
Mission DAG
The dependency graph of a mission's tasks: which are finished, which are running, and which are waiting on others. It is how a six-session mission stays legible instead of becoming six scrolling terminals.
Hand it the work.
Walk away.
macOS, Linux, and Windows. Around 13 MB. Free and open source.