OpenClaw hosting, explained
OpenClaw is the most capable open-source AI agent you can run yourself, and “run yourself” is exactly the part most people don’t want. Hosted OpenClaw means your agent lives on an always-on cloud machine that someone else sets up, secures, and keeps running. Here’s what that actually involves.
First, what OpenClaw is
OpenClaw is open-source (MIT-licensed) software that turns an AI model into a personal agent: something you message from Telegram, Discord, Slack, WhatsApp, or another chat app you already use, that can actually do things: work with files, browse, run scheduled jobs, remember you between conversations. It runs as a single always-on process called the Gateway on a computer you choose, and on its own it’s bring-your-own-key: the thinking is done by a model provider you pay directly, unless your host includes that usage in its plan.
That “computer you choose” is the whole question. An agent is only as available as its machine, and only as safe as its machine’s security. Run it on your laptop and it sleeps when you do; run it on a rented server and you’ve signed up to administer a Linux box running autonomous software. Hosted OpenClaw is the third answer.
What a host manages, and what stays yours
With hosted OpenClaw, the division of labor is clean. The host owns the machine and its operation:
The always-on computer
A private cloud machine that never sleeps, so your agent is reachable at 7am and working overnight.
Security
The gateway kept off the open internet, access that authenticates you, and a setup that’s security-checked: the whole checklist, maintained.
Backups
Your agent’s memory backed up automatically, with restores that have actually been tested.
Updates
Software versions pinned and upgrades handled deliberately. No surprise breakage from a moving “latest.”
And you keep the parts that make the agent yours:
- Your agent’s brain. OpenClaw works with all the major model providers, and hosting shouldn’t lock you into one. At Everpod, up to $10/mo of model usage comes included so there’s nothing to set up on day one, and running on your own provider key stays fully supported, with nothing added on top of it.
- Your agent itself. Its name, personality, memory, and workspace are yours, built through OpenClaw’s own setup. Hosted OpenClaw is still OpenClaw, with its native dashboard, not a lookalike wrapper.
- Your channels. You decide where it talks: Telegram, Discord, Slack, WhatsApp, and the rest. We walk you through the first one, and your agent can set up others when you ask.
What to look for in any OpenClaw host
Hosting an autonomous agent is security-sensitive hosting, and the questions worth asking any provider are specific:
- Is the gateway ever on the open internet? The only right answer is no; here’s why that rule is absolute. Access should go through an authenticated, private path.
- Do I share a machine with other customers? OpenClaw’s own trust model is one user per gateway, one boundary per machine. One customer per machine is the only hosting that keeps that boundary.
- What happens when I cancel? A clear answer covers the machine, the agent’s memory, and the backups. Deletion that leaves backups behind isn’t deletion.
- How is model usage billed? Whatever the shape, an included allowance or your own key, the answers you want are hard caps (no surprise bills), plan sizes in real numbers, and the freedom to switch to your own provider account. An offering that only works through its bundled usage couples your agent’s brain to the host’s pricing.
How Everpod does it
Everpod’s answer to those questions: your agent gets a pod, a private, always-on cloud computer of its own, one customer per pod, no exceptions. The pod accepts no incoming connections from the open internet unless you switch on a public address for your agent, and the control panel is never reachable through it; you reach your agent’s native control panel through a private, authenticated path, signing in with your email. Updates are applied for you after testing, backups run every day with restores tested, and its memory survives restarts and reboots. It’s $29/mo with up to $10 of model usage included every month, hard-capped so there are no surprise bills. Bigger plans start at $59 a month when you want more usage, and bring-your-own-key is supported if you prefer. Cancelling deletes the pod, the memory, and the backups 7 days after the paid period ends, a grace window in case you change your mind.