Skip to content
// Glossary · Virex Remote

Async coding agents

Also called: unattended agents, background agents, overnight runs

What are async coding agents?

Async coding agents are coding agents that keep working while nobody is at the machine. The work is started deliberately and then left to run, which means the agent will reach decisions with no one present to answer them. Whether async work is valuable depends almost entirely on how those blocking moments are handled.

// In more depth

What changes when nobody is watching.

The capability is not new. What is new is the failure mode.

01

Blocking becomes expensive

A run that pauses for an answer costs seconds when you are at the desk and hours when you are not. The pause is the same, the cost is entirely about who is present.

02

Silence is ambiguous

A session that has produced no output for an hour might be working hard or waiting for an answer. Without a visible pending state, those look identical.

03

Review moves later

You are not watching the work happen, so the record afterwards has to carry what happened and what was approved along the way.

// Why it matters

Why this is the interesting case.

Async is where oversight stops being a nicety.

Long jobs get used

Overnight work only pays off if a blocked run can reach somebody. Otherwise the job stalls in the first ten minutes and waits until morning.

Decisions travel

A permission or plan decision surfaced on a phone takes a minute to answer, which is what turns unattended hours into finished work.

Cost has to be predictable

Unattended is the worst case for usage-metered tooling, because nobody is present to notice. Virex bills a flat $15 per seat per month with no usage component.

// Commonly confused with

Commonly confused with.

Three things async work is not.

Autonomous agentsAutonomy is about how little direction an agent needs. Async is about whether anyone is present. An agent can be highly directed and still run unattended.
Hosted agentsRunning on vendor infrastructure is one way to do async work. Keeping the agent on your own machine and supervising it remotely is another.
Background jobsA background job runs to completion or fails. An agent session pauses for judgment, which is a different problem than retrying a failed task.
// FAQ

Async agent questions

Do I need my machine on for async work?

With Virex Remote, yes. The agent runs on your hardware, so the machine has to stay awake and connected. That is the trade for your code never leaving it.

How do I know a run is blocked rather than working?

Pending decisions surface as their own state rather than as text in a stream, so a blocked run is visibly blocked instead of just quiet.

Can I answer a decision from my phone?

Yes. Approving or denying a permission request or a plan decision works from any browser or device.

Does a long overnight run cost more?

No. Virex drives the interactive CLI, so model usage rides the Claude subscription on your machine and a seat is a flat $15 per month.

What if I want to stop a run mid-flight?

You can interrupt the current activity, or send an instruction into the session to redirect it.

Let the run reach you.

Virex Remote surfaces a blocked session wherever you are, so unattended hours turn into finished work.