Cloud agents

Running your AI workers from a phone

A terminal agent needs a terminal. A queued job needs a text field and a network.

The short answer

Assigning AI work from a phone is possible when execution happens on a server rather than on a device. Polaris ships a Flutter mobile app, live now as an installable web app, built around the copilot with voice input and spoken replies. Assigning a task from it writes the same queue row a desktop assignment writes, and the delivery comment appears on both.

Mobile app: Flutter, installable PWA
Input: Typing or voice
App Store status: Prepped, not released

Why this works at all

Nothing about a Polaris job runs on the device that started it. Assignment writes a row in Postgres, a trigger queues the job, and a process on a Fly machine claims it. The client is a viewer of a database, so any client that can write one row and subscribe to a few tables is sufficient.

That is the practical difference from a terminal agent. Reaching a shell from a phone means SSH, a keyboard you do not want, and a session that dies when the connection drops. Reaching a queue means typing a sentence.

What the mobile app does today

Described from the shipped app, not from a roadmap.

The copilot sits at the centre
The Chief of Staff is the primary surface rather than a list of projects, which suits a screen where you have one thumb and one intention.
Voice in, voice back
Spoken input and spoken replies are part of the interface, so briefing a worker on a walk is a normal use rather than an accessibility afterthought.
The same workspace
It is the same database and the same organisation as the desktop app, with the same tasks, buckets, comments and workers.
Installable now
The app is live as an installable web app you can add to a home screen without an app store.

What a phone is good and bad at here

Action	On a phone	Better on a desktop
Assigning a task to a worker	Fine
Briefing by voice	Better than typing
Reading a delivery comment	Fine
Watching a live session stream	Fine
Writing a detailed checklist		Yes
Reviewing a long document with comments		Yes
Configuring connections and credentials		Yes

What keeps running after you close the lid

The honest version: nothing on your laptop survives, so the work has to not be on your laptop.

Assigning a task to an AI worker

There is no prompt box. The task is the prompt, and the checklist is the contract.

Agents deliver, humans close

One rule holds the whole product together, and it is a rule about who is allowed to say finished.

Agent runtime

Not the model, not the framework: the thing that actually runs the job.

AI worker

The difference between an agent you prompt and an agent you assign work to.

Questions people ask

+Can I run a coding agent on my phone with this?

No. The Polaris runtime has no shell and no repository checkout, on any device. What a phone gives you is the ability to assign research, writing, analysis and review work to a worker and to read what came back.

+Do I need to install anything?

No. The mobile app is live as an installable web app, so a browser and an add-to-home-screen is the whole installation. A native store release is prepared but not published.

+Does voice input change what the worker receives?

No. Voice is an input method on the client. What reaches the queue is the same task title, description, checklist and comment text a keyboard would have produced.

+Will I be notified when a delivery arrives?

The app subscribes to task changes, comments and job events in realtime, so a delivery appears while the app is open. Push notifications to a closed phone are not part of the product today.

Cloud agents that keep working after you close the laptop

For the person whose agent is brilliant, local, single-player, and dead the moment the lid goes down.