Agent Workflows

WebTest AI is designed for agentic development without turning runtime truth into an open-ended prompt.

Recommended pattern:

Let humans or agents draft Markdown specs.
Run webtest-ai run for deterministic browser execution.
Use reports, screenshots, traces, and CI summaries as evidence.
Use webtest-ai report export for project-specific automation.
Use webtest-ai mcp when an agent needs local access to reports, specs, artifacts, or explicit browser debug tools.

Model-backed commands such as discover and recorder cleanup require explicit model configuration. They should produce proposals for review, not uncontrolled rewrites.