Test workspacePublic simulation
Usage guideNo model connected

Try the complete test workspace.

Inspect the sandbox, describe the expected outcome and start a run. Follow its live BPMN route, inspect assertions and evidence, add feedback, then edit and rerun a new version.

Only synthetic data can be tested here. Runs are private to this browser and expire after 30 minutes. A selected preview model performs real planning; the built-in workflow requires no model.

Request access and a demo

Run a real test in the sandbox.

Describe what you want tested. The agent discovers the application and controls, chooses relevant risk checks, and gathers evidence. Model and application settings provide optional context.

Ready to configure

1. Choose your model

Connect checks the selected model for a response.

2. Application context (optional)

The agent can find open automation tabs or infer a website from your prompt. A URL or selected tab is optional context.

Application observations
Choose a target, then inspect its available controls.

3. Define the check

Target action and evidence controls

The agent discovers controls automatically. Add selectors only when you want to give it a useful hint.

Password inputs are always masked. Add business-sensitive fields before collecting screenshots.

The agent starts permitted discovery immediately. Model proposals never execute as shell or Python code.

4. Review the evidence

The route appears as the agent works.

Every attempt has a place.

Start a test to see observations, policy decisions, retries and assertions connected in a live BPMN trace.

No execution events yet.