Sam's day with AssistOS

Sam writes books, coordinates marketing for his books and leads a research team. This morning, a chapter, a book launch, an experiment and a European project budget all need his attention. He already uses AI, but still spends hours finding files, repeating context and carrying decisions between tools.

In AssistOS, Sam opens WebChat, the interface to the Copilot, and describes what needs doing. Copilot coordinates the team's virtual workers, called Robots, and assigns their tasks. Books, Marketing and Research have separate team workspaces. Each workspace has its own RoboTeam, with a different set of Robots. Books has SCRIPTA, Proof and Layout. Marketing has Scout, Studio and Pulse. Research has Atlas for experiments, Vera for verification and Hugo for budgets. When Sam changes workspace, he works with that team's Robots and colleagues. Each project has its own AssistOS workspace and meeting rooms, with access limited to the people and Robots involved.

For the book, Sam builds a RoboTeam. SCRIPTA helps write, Proof reviews and Layout prepares the pages. He chooses agents from the AssistOS catalog, or asks his own IT team to prepare custom ones. Agents are software packages that run in Linux containers. A Robot is a configured instance of that software, with its own subscriptions, preferences, permitted skills and instructions. Sam also sets its role and workspace access. Two Robots can use the same agent package with different settings and responsibilities. Each Robot keeps its home and task history even when its underlying software changes.

Sam follows the Robots' background work in WebChat. Robots can control browsers, use desktop applications and run command-line tasks through their configured agents. These can include coding agents or tools that check sources and results. Sam can redirect, pause or stop the work, then return when there is something to review.

Sam opens the book in Explorer. This is AssistOS as an IDE, an integrated development environment. Software developers use one to keep their files, editor and checks together. Sam does the same for his book. He compares SCRIPTA's edits with the original chapter and checks Proof's notes before accepting the revision.

Sam makes the editorial decisions. When he approves the finished book, SCRIPTA publishes it on scriptahub.com.

He then switches to Marketing, where Scout researches publications in its browser. Sam can watch, take control and review the outreach before sending it.

In Research, Hugo prepares the European project budget in a desktop spreadsheet. Sam can observe that work too, and check the figures with the project coordinator before submission. In a separate AssistOS workspace for the study, Atlas compares baseline and smoothed outputs from large language models while Vera evaluates the results. Any workspace folder can be put in a GitHub repository or remain local. The researchers share their scripts through GitHub so colleagues can review changes, while keeping private notes on their own storage.

To discuss the findings, Sam joins his colleagues and the RoboTeam in WebMeet. The Robots are visible participants in the video call. A colleague asks Vera to add another test to the shared plan, and everyone sees the change. The room keeps the discussion and work within the study's approved context.

As Sam moves between teams, their work stays organized. This is the OS idea. An operating system manages programs and resources; AssistOS aims to manage virtual workers, their homes, tasks and permissions. It runs above existing operating systems. Teams keep their files and history when they change agents, models or where the work runs. A team can share an AssistOS installation on a private server or in the cloud. Colleagues connect from a laptop or phone, manage Robots together and follow each other's tasks across the subprojects they can access. Linux containers help isolate agent code and protect the machine hosting the workspace by limiting access to its resources.

Sam still writes, researches and makes decisions. His Robots take on more of the supporting work, so he spends less time moving information and more time guiding his teams. That is the AssistOS vision: useful work he can delegate, inspect and keep.

There is already code behind this vision. Axiologic Research developed AssistOS within Achilles, a European research project, as an environment for controlling AI agents. The platform is open source, and its source code is available on GitHub. The agents it runs can have different licenses, including proprietary code. Getting started takes a few setup steps, described on the Installation page. The Research page links to the technical documentation.