Build
Read what you are about to switch on.
Every route to a running agent — the studio, a template, the marketplace, the CLI, the API — passes through the same review screen. It states the permissions in full, in words, and it is the only place the deploy button exists.
One screen, every path
A marketplace install and a hand-written manifest produce the same review. There is no faster route that shows you less, because the faster route is exactly where a dangerous permission would slip through.
Written for the person accountable
"Can send email from your company mailbox", not "mail.send scope granted". The person signing off on an agent is often not the person who configured it, and the screen is written for the signer.
Changes are diffed, not restated
Re-deploying an existing agent shows what changed against the running version. A screen that re-lists forty unchanged permissions trains people to skip it.
You can dry-run from here
The review screen offers a dry run against your real recent history before you commit. Seeing what it would have done yesterday is more convincing than any description of what it will do tomorrow.
Agent Studio
The rest of this area.
Turn a described job into a configured agent.
Describe the job
Write the job in plain words. No canvas, no nodes.
What gets built
The five things a sentence becomes, all visible before deploy.
Guided creation
Five questions, in the order they actually matter.
Versions and rollback
Agents version like software, and roll back like it too.
Ownership
Every agent has a human owner. No orphans.
Give AI a job.
Build an agent, choose what it can access, define what it may do, and deploy it to work safely across your business.
No code required · Full SDKs and API · Stop any agent instantly · [email protected]