Compare
Keep the CLI. Add five more, and a conductor.
This is not a comparison against a rival. Loom is built on Claude Code, six real sessions of it, with a reasoning model planning, dispatching, recovering, and verifying on top of the exact CLI you already trust.
Same foundation
The real thing, not a wrapper.
Loom launches six genuine Claude Code CLI sessions in native PTYs, real TUI and all, authenticated by the Claude login you already have. No API key for the fleet, no forked agent, nothing to relearn. Watch any terminal, or click in and type when you want to steer by hand.
# Claude Code, installed the usual way
$ curl -fsSL https://claude.ai/install.sh | bash
# sign in once
$ claude
# Loom conducts that same login, six sessions at a timeWhat one terminal cannot do
The conductor's job.
Plan
Hand over a goal and a bring-your-own-key Conductor breaks it into a mission DAG of tasks, instead of one long prompt that drifts.
Dispatch and recover
Tasks fan out across six sessions. Loom answers permission prompts with only the safe affirmative, recovers rate limits and stalls, nudges drifting sessions, and rotates pooled Claude accounts on usage limits.
Verify
Done is checked before it counts. Live activity strips show what every session is doing, so a premature finish gets caught instead of shipped.
Common questions
For the CLI faithful.
Do I lose anything from the raw CLI?
No. Each terminal is a fully interactive Claude Code session. Shift+tab still cycles a terminal's permission mode, and auto-accept can be switched off whenever you want manual control. The permission modes guide covers the details.
Why six sessions instead of one?
Independent tasks stop queueing behind each other. A refactor, its tests, and its docs can run in parallel while the mission DAG tracks what depends on what.
What does this cost on top of Claude Code?
Nothing for the app. Loom is free, Apache-2.0, and about 13 MB. The fleet uses your existing Claude login, and the Conductor uses whatever key you bring, or a local model through LM Studio, MLX, or Ollama.
Hand it the work.
Walk away.
macOS, Linux, and Windows. Around 13 MB. Free and open source.