Use cases
Build the whole API surface at once.
Endpoints, validation, tests, and docs are four workstreams pretending to be a sequence. Loom dispatches them across six real Claude Code CLI sessions and verifies each piece before it counts.
The split
Scaffold, harden, document.
Scaffold
Sessions stand up resources and routes from your brief, one group of endpoints per terminal, with consistent error shapes across the whole surface.
Harden
Validation, auth checks, and edge-case tests land in parallel with the handlers they cover, not as a someday follow-up after launch.
Document
Reference docs and request examples are written from the actual code in the same mission, so they ship in sync with the endpoints they describe.
Mission control
Dependencies hold, stalls do not.
The Conductor plans the goal into a mission DAG, so schemas land before handlers and handlers before tests. It watches every terminal, answers permission prompts, nudges sessions that drift, relaunches any CLI that exits, and re-sends the task after recovery. When a session claims it is finished, the work is re-checked before the slot is marked done.
# typed once into the goal bar
$ Build the /projects and /invites REST endpoints:
CRUD routes, request validation, auth middleware,
integration tests, and reference docs with curl
examples. Match the error format used in /users.Questions
Common questions.
Which models do the work?
The fleet is six official Claude Code CLI sessions running on your existing Claude login, no API key needed. The Conductor that plans and dispatches is bring-your-own-key: OpenAI, Anthropic, Google, xAI, Groq, Cerebras, DeepSeek, Mistral, OpenRouter, any OpenAI-compatible endpoint, or local LM Studio, MLX, and Ollama. See models.
Can I poke the endpoints while the fleet works?
Yes. The built-in WebGL terminal supports tabs and splits with native PTYs, so you can curl your routes in a spare pane. Dev servers are auto-detected and opened in an inline web preview.
How do I review before pushing?
The source control panel shows staged changes, diffs, and a full git graph of what the fleet shipped, so you review everything like a pull request before anything leaves your machine. See reviewing fleet work.
Hand it the work.
Walk away.
macOS, Linux, and Windows. Around 13 MB. Free and open source.