Hand the job to AI agents.
Once, or every day.

No. The panel opens on Chat: you type what you want in plain language and the agent does it on its machine — browsing, writing files, running whatever it needs to. You never see a terminal unless you go looking for one. Everything it makes lands in Files, downloadable in a click.

No. Invites are gone — machines are open to every account, free tier included. Sign up, create a machine, and it is running in under a minute. The model hub, the panel, and the CLI are open too: install it and run `octomind login`.

A container that belongs to you, with octomind preinstalled, sudo and Docker. You choose the size (CPU/RAM), and on Pro you can bring your own image. It runs your sessions, keeps your files, and suspends itself when idle so it costs almost nothing while you sleep.

Idle machines bill memory only — $0.012 per GB-hour, about 2.4 cents an hour on Small — and after a few minutes of inactivity (5 min on Free, 15 on Pro — configurable) they suspend automatically; from then on you pay storage only. There is never a surprise: one live usage allowance is always visible in the panel.

Everything the agent uses, from one balance in dollars: the AI model’s work (what the provider charges us plus 5%), the machine’s time while it runs, storage, and any images, video or voice it produces. Your plan includes usage worth its price — Pro’s $50 a month is $50 of usage — and prepaid credits pick up after that, at the same prices.

Both. Drag, paste or attach an image in the composer and the agent looks at it — on any plan, at no extra cost, as long as the model you picked has vision (the attach button tells you before you send if it does not). Voice is on paid plans: hold the mic and your words arrive as text you can edit, and from Telegram or WhatsApp you can send a voice note and get the answer back as one. Listening costs $0.010 a minute, speaking $0.025, on top of the model — and spoken replies stay off until you switch them on in Settings.

No. Models come through the built-in gateway: log in once and every machine can use them. The free tier carries a curated roster; a paid plan opens the whole catalog — Claude, GPT, GLM, Qwen. Everything is billed the same way: what the provider charges us plus 5%, from your allowance and then from credits.

Sessions and your shared directory are durable — suspend never touches them, and every session replays in full when you come back. If you cancel Pro, nothing is deleted for 30 days — resubscribe within the grace period and everything is exactly as you left it. Your files are downloadable from the Files tab whenever you want them.

Yes, on Pro: build any image FROM our base (that guarantees octomind and the mount contract), push it to a public registry, and point the panel at it. We pull and verify — you get your exact toolchain.

Yes, on every plan. Everything the panel does is a REST API: create a machine, POST a prompt at an agent session, poll or stream the result, delete the machine. It uses its own developer keys (om_…), separate from model gateway keys — create one on the Keys page; the full reference lives on the panel Docs page.

Both, always. The panel is a plain conversation first; open Advanced at the foot of the sidebar for a full web terminal into the same container, machines, workflows and the config editor. Nothing is locked behind it — shells are per-connection, but your files and sessions always persist.