Run a fleet of coding agents on machines you own.
Bring your own CLI, your own subscription, your own hardware. They keep working when you close the laptop, reach you on your phone when they need you, and leave nothing to unpick if you walk away.
Nothing to pay Tecelon. You bring the machine and the model — and you keep both.
- It doesn't stop. Close the laptop; the container on your machine keeps going.
- It asks, you answer. Its questions arrive on WhatsApp. Reply yes, or say what to do instead.
- Ten at once. One list shows every worker, and which of them needs you.
What people do with it
-
Three ideas before lunch
One worker per idea, running at the same time. Judge them by what they built, not by the pitch.
-
Work on the fast machine
Put the job on the desktop under your desk instead of the laptop on your knees, and go out.
-
Ten agents, not ten tabs
A fleet, with one place that says what each is doing and where each is stuck.
-
No demo-to-production cliff
Every worker builds in containers, at a real address, from the first commit. The demo is the thing.
Three steps to your first worker
-
1
Connect a machine
Any computer with Docker — a desktop, a server, a spare Mac. One command, and it dials out: no port to open.
docker run -d registry.tecelon.io/tecelon/host-agent -
2
Say what to build
Type the job. The worker works out which repositories it needs, acts as you on GitHub, and opens the pull request.
-
3
Watch it, or walk away
Its terminal is a tab in your browser. Its questions are a message on your phone. Both, or neither.
Bring your own everything
Tecelon supplies the place agents run and the way you reach them. The rest stays yours.
- Your CLI. Claude Code today, on the machine's own install; other harnesses next.
- Your subscription or key. Anthropic, or anything that speaks its API.
- Your machines. A desktop, a server, a spare Mac — or one a colleague shares.
- Your GitHub. A worker acts as you, in your own repositories.
- Your login. You sign in on the machine itself; Tecelon never sees it.
Nothing to unpick
Lock-in is what you would have to undo to leave. Here there is almost nothing: it is your hardware, and what a worker makes is ordinary.
- Plain containers and a Dockerfile — they run anywhere, cloud or not
- The work lands in your repositories, commit by commit
docker exec -itinto any worker; it is a real CLI session, not a wrapper- Take the machine back at any time: stop the agent, keep everything on it
Not another hosted agent
Start with the machine you already have
Connect it, give a worker its first job, and close the laptop.