Zero to a graded verdict, step by step
No test code, no selector spelunking, no grid to provision. Describe what to test, run it in your own Chrome, read the evidence.
Six steps, one afternoon of margin
Create your workspace
Sign up with a work email and name your workspace. Everything you make — tests, runs, playbooks — is stored per-workspace and isolated from every other tenant.
Install the browser extension
The agent drives your own Chrome, so real logged-in sessions, VPNs, and SSO all just work. Install the extension, and the browser you already trust becomes the test runner.
Connect your app and docs
Point TestAutomate at your app's URL and ingest your help center or internal docs. They become retrievable playbooks — the agent learns your flows before its first run. See how knowledge works.
Generate a draft suite
Describe your critical paths in plain English — type them or dictate by voice. Each becomes a portable YAML test you can read, edit, and version. Details on authoring.
Run and read verdicts
Each run is graded pass, fail, or blocked, with screenshot evidence at every step. Failures are re-driven by a stronger model before you ever see red. Details on verdicts.
Connect Jira
Confirmed failures become deduped Jira bugs with evidence attached — confirm-gated, never automatic. Point a sweep at your board and every done ticket gets verified. See the bug loop.
What a good first test looks like
Start with the flow that would page you at 2am if it broke. Keep it to one user intent per test.
- State the intent, not the clicks. "Sign in and create an invoice for a new customer" beats twelve UI steps.
- Say what done looks like. "The invoice appears in the list with status Draft" gives the verifier something concrete to grade.
- Let blocked be blocked. If staging is down, the run is blocked — not failed. Broken infrastructure never counts as a product bug.
Where to go next
- Execution — how the agent drives your real browser, and why accessibility-tree-first reading keeps runs fast and cheap.
- Economics — BYOK setup for Anthropic, Bedrock, Gemini, or Groq; live meters; hard budget ceilings.
- Sprint UAT with Jira — point a sweep at a board and get a graded report per ticket.
- The sprint UAT guide — a longer walkthrough of running verification sweeps every sprint.
Your first graded run is one plain-English test away
Plain English in, evidence-backed verdicts out.