Agents & Sessions
How agents execute work in isolated browser sessions and what artifacts you get back.
Certyn executes work using AI agents running inside isolated sessions.
What a session is
A session is the runtime unit of work:
- a fresh, isolated browser environment
- an agent that reads the assignment (task + tests + environment + wiki)
- artifact recording (logs, screenshots, trace)
Sessions do not share state with each other.
How the agent executes
At a high level, each session loops:
- observe the page state
- decide the next action
- execute (click/type/navigate/scroll)
- verify expectations
The goal is resilience to minor UI changes without maintaining brittle locator code.
What you get back
Every session produces evidence you can use for triage and verification:
- action timeline
- screenshots
- agent trace
- final outcome (pass/fail/error) and reasoning
Live viewing
While a session is running, you can open the live view (for example via Watch Live on a process page or the issue live route) to see the real browser the agent is controlling.
