Examples — what people run on GridRocket

Six setups, the machine each one uses, and what it looks like once it is live. Every one of them is the same product — an agent on its own machine behind an HTTPS address.

An always-on coding agent

You want an agent that can pick up a task, work through a repository and keep going while you are in meetings — not one that dies when your laptop sleeps.

Machine: A coding agent on a 2 CPU / 4 GB machine, in the region closest to you.
Once live: You open the machine address, paste the access token, and the agent's workspace is there. SSH in when you want to inspect the checkout directly.
Outcome: Work continues without your laptop. Sessions survive closing the browser tab.

A research and browsing agent

You need an agent that opens pages, follows links and collects findings — the kind of work that hammers a laptop's memory and battery.

Machine: A browser-capable agent on a 2 CPU / 4 GB machine; step up a size for heavy parallel browsing.
Once live: The agent runs behind its own HTTPS address with a gateway sign-in. You watch it work from any device.
Outcome: Long research runs finish on their own and the results are waiting when you come back.

Overnight scheduled automation

A job needs to run every night: pull data, summarise it, file the result. It has to happen whether or not anyone is awake.

Machine: A small dedicated machine you never turn off, in whichever region the data lives.
Once live: You SSH in once to add the schedule, then leave it. Health checks confirm the machine is still answering each hour.
Outcome: The job runs on its own and the console tells you if the machine ever stops responding.

A private assistant with your own API keys

You want an assistant using your own model keys, without those keys sitting inside somebody else's multi-tenant service.

Machine: Any agent on a machine that is yours alone — no shared runtime, no shared disk.
Once live: Your keys live on your machine's disk. The only public entry point is the HTTPS address, protected by the access token.
Outcome: One tenant per machine, so nothing about your assistant shares space with anyone else.

A shared endpoint for a small team

Three or four people want to use the same agent instead of each running their own copy locally and disagreeing about state.

Machine: A mid-size machine — more CPU and memory, since several people hit it at once.
Once live: Share the machine address and the access token with the team. Rotate the token by relaunching the machine.
Outcome: One agent, one address, one bill — instead of four half-configured local setups.

Long-running jobs that outlive a laptop

A migration, a bulk transform or a multi-hour agent run that must not be interrupted by a reboot, a commute or a dead battery.

Machine: Size chosen for the job; destroy the machine when the job is done and relaunch later within the same paid month.
Once live: Start it, close the lid, check the provisioning log and health history from your phone.
Outcome: The job finishes because nothing it depends on is on your desk.