skip to content
~/docs~/integrations~/faq

</aethereum> / for / operators

Operators

Run agents like a team you manage, not tabs you babysit

An operator does not want to watch agents work; an operator wants the work done overnight, inside a budget, with anything risky waiting for a yes in the morning. Aethereum turns that into one flow: pick a standing mission (nightly test fix, dependency patrol, code review, triage), and the scheduler queues managed runs that execute on your machines or your CI under guarded autopilot. The full transcript never leaves the machine that ran it; you read the verified summary and tap approve.

where it breaks

Autonomy without a leash

An unattended agent can burn budget or push something irreversible; a supervised one saves you no time.

Cloud agents want your repo

Hosted agent products clone your source into their sandbox; that is a non-starter for client work.

Ten agents, no roll-up

Work happens in scattered terminals with no schedule, no costs, no approvals, and no morning summary.

how it works
  1. aethereum mission start nightly-test-fix seeds the brief, the goal, and the scheduled auto tickets in one command.
  2. Enroll a machine with aethereum listen --enable, naming whose tasks it may run.
  3. Runs execute overnight under guarded autopilot: injection-scanned, budget-killed live, irreversible actions parked.
  4. Read the morning report in the control room; approve or reject from the approvals dock.
npx aethereum init

Guarded autopilot is opt-in per machine per room, and nothing executes until the machine's owner allowlists the task authors. The server sees run summaries and token counts, never your source. Everything is free right now.

get started free →what managed runs arehow guarded autopilot worksvs Gemini Managed Agents