Request a demo

Agent OS · Architecture

Four layers, one governance plane.

The page for the architect in the room. Everything here is shared by every application, which is why the security review happens once.

Under all four

  • Audit trail
  • Version control
  • Access
  • Evals
  • Credentials

What sits in this layer

Module 3 · 3.2.SLive
  • Editor · long documents
  • Tables · bound to source
  • Forms · regulated fields
  • Dashboards · state
  • Chat · where it earns it

Your team and the agents work in the same surface, on the same permissions model.

Data secure in your cloud21 CFR Part 11-awareSOC 2 Type II (in progress)ISO 27001 (in progress)Your data never trains foundation modelsTrust and security →

Governance

Every one of these is a property of the stack itself.

Which is what makes them true of an application nobody has built yet.

  • 01

    Audit trail

    Every run logged, and every authored value linked to the source page behind it, so a review starts from evidence rather than from a search.

  • 02

    Version control

    Documents, workflows and configuration versioned together, so a change can be shown as it was made rather than reconstructed afterwards.

  • 03

    Access

    Role-based access and enterprise SSO across every application, so adding a function does not mean adding an identity story.

  • 04

    Evals

    Every workflow carries its eval set, and a change does not ship past a regression.

  • 05

    Credentials

    Brokered at the boundary and never sitting inside a workflow, so a workflow cannot carry a secret out with it.

  • Deployment details, data residency and certification status live on Trust and security.

    The full trust posture

Under the hood

Every application is assembled from the same six components.

Describe a workflow in plain language and Agent OS assembles it into a typed data model, agentic workflows and real screens. This is the one section on the site that uses our own names for things.

  • Agent runtime

    SmartRunners

    Serverless multi-agent flows: tool loops, long-horizon agents, model factory. Long-running agents hold no infrastructure between steps.

  • Data and audit

    System of record

    A Postgres system of record, plus every artifact a workflow produces. Nothing an agent writes exists outside the record.

  • Interface

    Component library

    An AI-ready editor, tables, forms, dashboards and chat, assembled into the application's screens. Your team reviews inside the application, on the same record the agents wrote to.

  • Integrations

    MCP-native

    SharePoint and GitHub live today, and any standard connector in about a week. Documents stay where your team already keeps them.

  • The builder

    SyncApp Agent

    Turns plain intent into a running application: entities, storage, workflows, screens. It is also how your own team extends one after handover.

  • Evals

    Non-negotiable

    Every workflow is gated by evaluations before release, and the gate stays for every change after it. A model swap has to earn its way in.

Want the deeper walkthrough?

We'll take your architects through the OS on a live deployment.

Talk to us

Your data stays in your environment. Customer data never trains foundation models.