Serverless that remembers, and keeps working.
Long-running AI workers that keep their state between runs and can act on a schedule.
Agents are in preview: built and available, not yet proven in production. Tell us what you’d run on one.
Part of a Solution. Built, released and secured together with the rest of your app: one place, one release.
What it does
What you get with Agents
Serverless that remembers
Each agent keeps its own state between runs; no database to wire up.
Works in the background
It can wake itself on a schedule, poll, follow up and carry on without being asked.
One agent per job
Run thousands (one per customer, per ticket, per deal), each isolated, and retire them when done.
Your model or ours
Use Waymaker One, or bring your own model key.
Use cases
What people build with it
One agent per deal
It remembers every touchpoint, chases the follow-up on day three, and retires when the deal closes.
One agent per support ticket
It triages, drafts a reply, waits for the customer and picks up where it left off.
A watcher that acts
Wakes every hour, checks what changed, and does something about it without being asked.
How it’s charged
Metered on run time, plus AI usage, drawn from your plan’s credits today.
See Host pricingThe technical detail
Commands, limits and step-by-step setup for Agents live on the Docs site.
Read the docsQuestions
How is an Agent different from an Ambassador?
An Ambassador wakes, does one piece of work and forgets. An Agent keeps its own state between runs and can wake itself on a schedule, so it suits work that unfolds over days, such as following up on a deal.
Are Agents production-ready?
Not yet. Agents are in preview: built and available, but not yet proven in production. We label that plainly so nobody discovers it the hard way.