Built for humans and agents
Humans and coding agents author the same reviewable tests, backed by your repository, skills, and MCP.
People and coding agents work with the same test files. A test describes the expected behavior in natural-language steps inside YAML, so anyone reviewing a change can read what it verifies.
The illustration below shows both authoring paths. People contribute their product knowledge; agents use repository context, rules, skills, and MCP to produce the same kind of reviewable test.
For humans
Describe a user journey and the outcome that matters. Each test has a stable ID, a name, a configured target, and a list of steps. Add context when the agent needs to understand a product term or the initial state of the app.
You can edit these files directly or use the dashboard. Both the dashboard and CLI read the same file-backed definitions. Start with your first test, then review the steps alongside the code they cover.
For agents
Give your coding agent the requirements and repository context. The agent quickstart connects it to a local agent-qa installation, while the packaged skills guide authoring, execution, and result triage.
Through the CLI or local MCP server, the agent can discover the workspace, author and validate tests, execute them, and inspect the resulting evidence. The public vostride.com MCP supplies documentation and setup guidance; execution happens through your local installation.
Follow the coding-agent verification loop to turn a code change into a focused test, an evidence-backed diagnosis, and a rerun.
One review process
Tests stay in your repository regardless of who writes them. Review the expected outcome against the requirement, validate the definition, and inspect the actual run result. Successful clicks alone do not establish that the behavior is correct.
Keep tests, configuration, and useful memory under the same version-control workflow as the rest of your project.