Build

Describe the job. Build the agent.

Write what you want done in the same words you would use to brief a new starter. The studio turns that into an agent with a trigger, a set of tools, a permission list, an escalation rule and a named owner — and shows you all five before anything runs.

Build an agent See how it works

You write a sentence, not a flowchart

No canvas, no nodes to drag, no branching diagram to maintain. A schedule, a condition and an escalation fit in one sentence, which is how people actually describe work.

Nothing is deployed from a description alone

The studio proposes; you approve. Every agent passes through a review screen that states its permissions in full, and that screen cannot be skipped — not by a template, not by the marketplace, not by the API.

The structured version is the real one

Your sentence is not the agent. It produces a manifest with explicit tools and grants, and that manifest is what executes — so the agent's behaviour does not drift as a model changes underneath it.

In this area