Set a pass rate.
Autopilot delivers it.
Give zTester a URL, 2 user journeys, and working auth. It discovers your app, generates deep multi-page tests, validates them, runs the full suite, and loops through repair until your target pass rate is hit — or tells you exactly why it can't.
curl -X POST https://ztester.zavecoder.com/api/v1/projects/$ID/autopilot \
-H "Authorization: Bearer $ZTESTER_API_KEY" \
-d '{
"baseUrl": "https://staging.myapp.com",
"targetPassRate": 80,
"journeys": [
"User signs up → verifies email → sees dashboard",
"Admin creates invoice → sends to customer → marks paid"
],
"authStrategy": { "type": "cookies", "cookies": [...] }
}'The quality loop
zTester doesn't just generate tests — it runs a full improvement cycle until the suite is trustworthy.
Give it context
A URL, 2–5 numbered user journeys, and working auth. Magic-link apps use session cookies. The readiness gate blocks discovery until inputs are strong enough to produce reliable tests.
Discovery generates deep tests
ActionExplorer clicks buttons and observes real state changes. The journey generator stitches multi-page funnels (tickets → cart → checkout → confirmation) from your brief. Tests start as drafts.
Validation promotes drafts
Each draft runs a probe execution. Only tests that pass become active. No silent accumulation of broken tests.
Full run checks the real pass rate
After validation, Autopilot runs the entire active suite against your app and measures the true pass rate — not just the validation probes.
Repair or rediscover until target hit
If pass rate < target: failing tests get AI-repaired (selector fixes, assertion updates), then focused rediscovery fills coverage gaps. Loop repeats up to maxIterations.
Autopilot session phases
Journey Tests
Deep multi-page tests, not smoke checks
Per-page tests tell you a button exists. Journey tests tell you a customer can actually buy a ticket. zTester synthesizes full purchase funnels, booking flows, and onboarding sequences from your brief — 13-16 step tests that traverse your real app from start to confirmation.
Everything the suite needs to stay healthy
Built for teams who need E2E coverage they can actually trust in CI.
Autopilot API
One POST call orchestrates the full cycle: readiness check → discovery → validation → CI run → repair → repeat until pass rate hits target.
Journey Tests
Multi-page end-to-end tests from your user journeys. Ticket purchase, booking flows, onboarding sequences — 13-16 step funnels with real business assertions.
Regression Detection
Every run is compared to the previous one. Regressed tests (pass → fail) and recovered tests (fail → pass) are surfaced immediately in the dashboard and webhooks.
ActionExplorer
Clicks buttons, fills forms, submits data, and observes real state changes. Generates 5-8 step workflow tests with success toasts, table updates, and validation errors.
AI Self-Healing
When selectors break after a UI change, AI repairs them automatically. The quality loop runs nightly: archive flaky tests, repair degraded ones, rescue stuck drafts.
Auth-Aware
Form login, OAuth, magic-link cookies, Playwright storageState. Detects silent auth failures on SSO/magic-link apps and tells you exactly how to fix them.
GitHub & Bitbucket
Connect your repo and zTester reads routes, entities, and form schemas from source. Discovery is richer and avoids hallucinated selectors.
MCP Server
Wire zTester into Claude Desktop or Cursor. 16 tools for the full workflow — bootstrap, discover, validate, run, and report failures — via natural language.
Quality Loop
Nightly cron: rescue stale drafts, archive 5× failures, AI-repair 2–4× failures. The suite improves autonomously without human intervention.
Use from anywhere
Ship with a suite you can trust
Set a target. Autopilot builds the suite and keeps it passing.