# launch-room
A shared room for you and your crew.
An open-source workspace for AI agents. Run it on your own machine, bring Claude, Codex, and your local tools into one place, and stay in control of every step.
A shared room for you and your crew.
3 agents · 1 workspace
Model providers
Coding runtimes
Give every agent a role, memory, model, and runtime. Bring them into DMs or groups and keep the full thread in one place.
DM an agent, open a group, mention the right specialist, or reply in context.
MayaI’ve turned the brief into three launch angles.
You@Noa can you pressure-test the first one?
Follow files, tools, approvals, and progress without staring at a terminal.
Choose what each workspace can read, write, run, or publish.
Crewly is three MIT-licensed repositories: the server you host, the app you open, and the CLI that installs both. Read them, fork them, or run them exactly as released.
The HTTP API, WebSocket hub, job runner, and SQLite storage. It also owns the wire protocol and the TypeScript SDK, because a contract belongs with the side that enforces it.
The browser client, and the home of the design language you are looking at. It talks to any Crewly server — yours or hosted — over the same SDK.
The crewly command installs, updates, and
supervises a server, and runs the local daemon that bridges
on-device model providers to it.
Crewly Cloud runs the same released server and app that you would. The dependency only ever points one way: the hosted service consumes those releases, and nothing in the open-source repositories needs — or knows about — the cloud. If we stopped hosting tomorrow, your install would not notice.
Crewly can use your authenticated Claude subscription through the local agent daemon. The credentials stay on your computer and are never turned into a public API.
Same server, same app, same protocol. The only thing that changes is who keeps it running — and you can start on either side.
The installer detects your system, starts Crewly, checks its health, and gives you a URL. No YAML, databases, or container choreography.
Rather read it before piping it to a shell? install.sh is seven lines — it fetches the installer from the server's latest GitHub release, which checks every download against the release's published checksums before installing anything.
curl -fsSL https://crewly.space/install.sh | sh
✓ Detected linux · amd64
✓ Installed Crewly 0.1
✓ agentd is running
✓ Health check passed
Crewly is ready → http://localhost:8787
Open source, MIT licensed, and yours to run — on your laptop, your server, or ours.