Skip to content

Deploy

Four reasons an agent starts working.

Something happened, a time arrived, a person asked, or another agent needed help. Each is stated in words on the agent's profile — "weekdays at 08:00", not a cron expression — because the person accountable for an agent should be able to read when it runs.

Why an agent starts

Stated in words on every agent's profile, never as a cron expression.

Something happened

A new lead arrivedAn invoice fell overdueA complaint was logged

A time arrived

Weekdays at 08:00Every FridayMonth end

A person asked

Run this now

Another agent needed help

Research this companyDraft this proposal

Something happened

A new lead, an incoming message, an invoice falling overdue, a complaint, a campaign crossing a threshold, an anomaly on your website. The agent starts because the business changed.

A time arrived

Every weekday at eight, every Friday, at month end. Written as words and shown as words, including on the run record.

A person asked

Run this now. Useful far beyond testing: an agent that can be invoked on demand is one somebody reaches for rather than working around.

Another agent needed help

A delegation from an agent in the same workspace, recorded as part of one chain — and under the specialist's own permissions, never the caller's.

Questions this answers

The things you currently have to ask somebody.

Each of these is a question an owner asks regularly and currently answers from memory, a spreadsheet, or by interrupting whoever is nearest.

When does this agent actually run?

What event started this particular run?

Can I run it now to see what it would do?

Which agent asked this one to do something, and why?

Where this stops

What it does not do.

A trigger is not permission to act. An agent woken by an event still has only the capabilities it was granted, and an event it has no permission to act on produces a run that reads the event and stops. Events are also not guaranteed to be instant — the platform is built for work measured in minutes, and an agent is the wrong tool for something that must happen within a second.

Next

Where this connects.

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]