For developers & SDETs
A guided path for developers and SDETs — gate regressions before merge, drive BugBrain from your IDE with the MCP server, automate with the API, and export real Playwright tests.
This track gets a developer or SDET wired into BugBrain: catch regressions before merge, drive it from your editor, and keep everything in code review.
Your goal
A pre-merge quality gate on your pull requests, BugBrain reachable from your IDE and CI, and exportable tests you own.
Your setup checklist#
Create a project and an API key
Add your app, then mint an API key. Use the PR/CI purpose for a least-privilege key.Connect the MCP server
Add BugBrain to Claude Code, Cursor, or Codex so you can run tests and read results without leaving your editor. See MCP server.Author or generate test cases
Write cases in plain English, generate them from a description, or let the agent author them from the app graph. See Test cases.Turn on PR automation
Install the GitHub App and link your repo so BugBrain tests the impacted flows on each pull request. See PR automation.Export Playwright when you need it
Any test case compiles to portable Playwright TypeScript you can run anywhere.
How execution stays robust#
BugBrain test steps describe intent, not selectors. At run time each step resolves through a four-tier fallback — a cached element fingerprint first, then semantic locators — so a renamed class or moved button doesn't immediately break the test. When the UI shifts, the agent heals the case and records the fix. Read the full model in self-healing tests.
Automate it#
- MCP — the fastest path for interactive use in your IDE. Scope-gated tools mean a key only exposes what it's allowed to do. See MCP tools.
- REST API — for CI pipelines and scripts. Authenticate with an
sk_live_…key. See the REST API reference. - Webhooks — subscribe to
run.completed,run.failed, andpr.check.completedto react in your own systems. See webhooks.
Least privilege by default
API-key scopes are additive permissions — an empty scope set means full access, so always pick the narrowest purpose preset that works. The privilege-sensitive scopes (key management, webhook writes) are enforced server-side.
Next steps#
Frequently asked questions
Does BugBrain lock me into a proprietary test format?
No. Test cases are stored as portable semantic steps and compile to exportable Playwright TypeScript you own. The MCP server and REST API give you programmatic access to everything.
How do I run BugBrain in CI?
Create a least-privilege API key (the PR/CI preset), call the REST API or use the MCP tools from your pipeline, and connect the GitHub App for advisory pre-merge checks. See PR automation.
How does it avoid flaky selectors?
Test steps express intent ("click the checkout button"), not brittle CSS paths. At execution they resolve via a cached fingerprint, then fall back through semantic locators — the self-healing model.