Getting started
Quickstart
Go from an empty workspace to a passing, reported test run.
1. Create a project
A project is your top-level container, usually one per product or repository. From Projects → New Project, name it and, if you use Jira, attach the project key so tickets and test cases can link automatically.
2. Connect at least one source of truth
Go to Settings → Integrations and connect whichever you already use: Jira for tickets, Figma for designs, Postman for API collections. None are required to start, but each one removes a step later: generating a suite from a real ticket beats describing it from scratch.
3. Generate your first suite
From Generate, either pull a ticket from your connected tracker or upload a PRD/spec document. TestCatalyst drafts a full suite of structured test cases from it. Review what it produced (edit, delete, or add cases); this is the AI doing the first draft, not the final word.
4. Run it
You have three ways to execute a suite, and you can mix all three over time:
| Method | Where | Good for |
|---|---|---|
| Scripted / agentic run | Runs → Run Tests | TestCatalyst's own AI drives a real Playwright browser against your suite, locally or on BrowserStack. |
| MCP run | Claude Desktop / Claude Code | Your own agent drives the test, using TestCatalyst's MCP server to report results. See MCP server. |
| Manual | Suite detail page | Walking through cases yourself, marking pass/fail as you go. |
Every run, scripted, agentic, or MCP, gets a live page with steps, screenshots, and pass/fail counts updating in real time. Share that URL instead of narrating results in chat.
5. Check the bug it filed
If a step fails, a bug sheet is created automatically with severity and repro steps pre-filled from the run. Open Bugs, adjust anything that needs a human judgment call, and push it to Jira or Trello when it's ready. TestCatalyst pings the right engineer on Slack or Teams the moment it's pushed.
6. Re-run and watch the bug close itself
Once the fix ships, re-run the same suite. When the linked test case passes again, the bug resolves on its own; nobody has to remember to go close the ticket.
7. Generate a report
From Reports → Generate, pick a type (Comprehensive, Exploratory, Regression, or API Test), and send it by email, export it to Confluence, or download it as a PDF. It carries your workspace's logo and colors if you've set up branding.
What's next