Understanding Runs

When you start a run (for example by executing a process), Certyn spins up isolated agent sessions and records evidence so you can review outcomes quickly.

Your first run

There is no special onboarding run. Your first run is a normal run: you ask for it in Ask Certyn (for example explore the checkout) or start a process. Certyn uses it to learn your app: it records what it learned in memory and writes test cases for what it finds. Before that, Certyn can check that it reaches your app and signs in, on a Machine you can watch.

Execution flow (high level)

1) A run is created

Certyn creates a parent record for the run to preserve auditability. This run ties together:

• what was requested (process, tags, environment)
• who/what triggered it (manual, schedule, CI)
• links to created sessions and artifacts

2) Sessions start

For each unit of work, Certyn creates a session: an isolated environment with a browser and agent runtime.

Sessions are isolated from each other (no shared cookies, storage, or filesystem state).

3) The agent executes

Inside each session, the agent:

1. reads the assignment (task + tests + environment context + wiki)
2. navigates the target environment
3. interacts with the UI (click/type/scroll/navigate)
4. verifies expected outcomes

4) Evidence is recorded

During execution, Certyn records artifacts like:

• step-by-step logs
• screenshots
• the agent trace (what it observed and why it made decisions)

5) Outcomes become Issues (when needed)

If a test fails or exploration finds something suspicious, Certyn can create an issue with reproduction context and evidence. Issues are designed to be easy to triage and verify.

Watching sessions live

While a session is running, you can watch it live from the UI (for example from a process page via Watch Live) or by opening the live route for the related issue or run.