AI agent hosting without the Mac mini
Your octo works on a managed machine: real Linux, Docker inside, a browser and a disk that keeps its files. It wakes on a message or a schedule, does the job, and goes back to sleep, so the hours it has nothing to do cost storage, not compute.
Every weekday at 7:00, check zapier.com/pricing and n8n.io/pricing and tell me only what changed.
Nothing to install — working in 60 seconds. Cancel anytime.
How it works
-
Say the job, skip the setup
Type what you want in the panel, Telegram, Slack or WhatsApp. Your first octo finds itself a machine, and models come through the built-in gateway, so there is no server to rent and no API key to paste.
-
AI agent hosting on a real Linux machine
It works in its own folder with sudo, a Docker Engine and, on the default image, a headless browser. It installs what it needs, runs code, and keeps its files on the machine between runs.
-
The result reaches you, then it sleeps
The answer lands in the chat you asked from. After an hour with nothing to do the machine suspends, and the next message or routine wakes it with its disk exactly as it was.
Jobs to hand over
Watch two pricing pages on a schedule
Every weekday at 7:00You say:“Check zapier.com/pricing and n8n.io/pricing. Compare them against what you found last time and tell me only what changed — prices, plans, wording. If nothing changed, say so in one line.”
What comes backThe first run saves a baseline; every run after reports only the changes, or one line saying nothing did.
Put a page online from its machine
OnceYou say:“Build a clean one-page landing page for Clay Hours, a weekend pottery studio in Lisbon: a headline, three benefits and a contact form. Put it online on a public link and send me the link.”
What comes backA live link served from the machine, checked for HTTP 200, for $0.04; the form validates but does not send.
See the real result →Turn the machine into a dev box
OnceYou say:“Set up this machine for Python development, create a small example web app and run it. Show me it works and what to try next.”
What comes backPython installed, a FastAPI todo app with 3 of 3 tests passing, running on a live URL with its interactive docs.
Run your docker compose stack
OnceYou say:“Clone my repo, start it with docker compose and give me a private link to the app so I can click through it.”
What comes backThe stack running in the machine’s own Docker, behind a link only your signed-in account can open.
A morning brief before you are up
Every morning at 8:00You say:“Find what is new about AI agents and automation tools since yesterday and write me a brief: the three things that actually matter, one line each, with links. Skip anything I would already know.”
What comes backThree lines with links in your chat; the real run cost $0.37.
The machine is right there on the octo’s page
This is a real octo’s Body tab: a Medium machine with 2 vCPU and 4 GB of RAM, the folder the octo works in, and the octos it shares the machine with. Open a shell here drops you into a terminal in that folder. The page linked below was built and put online by an octo, for $0.04.
Open the page an octo built and put online →
What it doesn't do
It is not on 24/7: it sleeps when idle
A machine suspends after 1 hour with nothing to do and wakes on the next message, routine or shell, so scheduled jobs still run on time. You can lengthen the idle timer per machine, but every warm hour bills memory ($0.012 per GB-hour). Read more →
Anything left running stops when it sleeps
Suspend stops the machine, so a dev server, a bot loop or an open shell ends; files on its disk and every conversation stay. Put repeating work in a routine instead of a loop that has to stay alive.
After 3 days asleep, waking takes minutes
A machine suspended for 3 days moves to cold archive: all its data is kept, but the next start restores it in a few minutes instead of seconds. Expect a slow first reply after a long break.
No GPU and no local models
Machines are CPU containers, from Small (1 vCPU, 2 GB) to Large (4 vCPU, 8 GB), and models run through the gateway at the provider’s price plus 5%. To run Ollama or LM Studio on hardware you own, self-host open-source octomind. Read more →
The cloud starts at Pro, and Large needs Max
Free is the model hub only, with no machine. Pro is $50 a month ($20 the first month) with 3 machine slots on Small and Medium; Large machines start on Max at $200. Read more →
Questions
What is AI agent hosting?
A place for an agent to run while your laptop is closed: a computer with its tools and files, and a way to reach you when it is done. Here each octo works on a managed Linux machine with Docker and a disk that keeps its files, and model calls go through a built-in gateway, so there is no server to rent and no key to wire up.
Do I need a Mac mini to run an AI agent?
No. A Mac mini makes sense if you want local models, macOS-only apps or hardware you own, and octomind is open source, so you can run it there yourself. If you want the job done and the answer in Telegram, Slack or WhatsApp, a cloud machine skips the purchase and the box on your desk.
Can my agent run 24/7?
Not by default, and on purpose. The machine suspends after an hour with nothing to do and wakes on a message, a routine or a shell, so scheduled jobs still run on time. You can lengthen the idle timer per machine, but every warm hour bills memory.
How much does AI agent hosting cost while the agent is idle?
Asleep, a machine bills storage only: 10 GB of its disk is included and more is $0.10 per GB-month. Awake but idle, a Small bills memory, about 2.4 cents an hour, and a Small kept warm all month is $17.52 of Pro’s $50 allowance.
Can I bring my OpenClaw or Hermes agent over?
Yes. Upload an export and its memories, instructions and skills come over; from OpenClaw or Hermes its scheduled jobs and Telegram bot come too. You preview what to import, and API keys are never imported.
Are several agents on one machine isolated from each other?
No. Octos on one machine share a user and a container; each keeps its own folder, but a folder is not a security wall. Give an octo that holds a client’s credentials a machine of its own.
More from Octomind Cloud
- Personal AI agent Your own AI agent for each thing you run, reporting back in Telegram, WhatsApp or Slack
- No-code agent builder Say the job in one sentence and get an agent with a cloud machine, memory and schedule.
- Competitor monitoring Checks rivals' pricing and feature pages on your schedule and tells you only what changed
- Telegram AI agent Text your octo in Telegram: it does the job on its own computer and replies there
Reading
Sixty seconds to your first octo.
Subscribe and hand it the job — $20 your first month, cancel anytime. Nothing to install, so nothing to uninstall.