qracle.ai
Sign in Request access

Private beta · 2026

QA that scans everything,
then thinks.

qracle runs ten specialist scanners in parallel against your app, then an LLM Oracle adjudicates every finding into one report — with a confidence score, blast radius, and a paste-ready prompt for your IDE-AI.

Request beta access View a sample report

Multi-LLM adjudication · BYOK or platform-managed · SARIF + Markdown out

What qracle scans

Ten specialists, one adjudicated report.

Each scanner wraps a mature OSS tool and runs in parallel. No single tool is the source of truth — the Oracle cross-validates and adjudicates every finding.

functional

User journeys

Playwright runs deterministic journeys; Stagehand handles agentic flows on staging.

a11y

WCAG conformance

axe-core walks every reachable view; narrow LLM-assist for alt-text and label ambiguity.

perf

Field + lab metrics

Lighthouse CI + Unlighthouse over the page tree, k6 for synthetic load smoke tests.

visual-diff

Pixel regression

Argos snapshots with pixelmatch diffing; Lost Pixel fallback when self-hosting is off.

security-zap

Dynamic + static

Authenticated OWASP ZAP DAST, read-mode by default; active-scan opt-in per surface.

supply-chain

Deps + secrets

Trivy on filesystem + image, gitleaks across git history. CVEs and tokens before they ship.

architecture

Drift detection

Layering violations, dead code, and coupling that crosses module boundaries.

broken-link

Reachability

Crawls internal + external links; flags 4xx/5xx and redirect chains across the surface.

ai-slop-syntactic

Hallucinated code

Undeclared imports, phantom APIs, and other syntactic tells of unreviewed LLM code.

pipeline-contract new

Workflow contracts

Consumed-but-not-produced artifacts, fixture imports in prod paths, empty auth refs, vendored-contract drift.

How it works

One report, not ten.

  1. 1

    Scan

    Point qracle at a URL (and optionally a repo). One Temporal workflow fans out all ten specialists in parallel — Functional emits a HAR that feeds the API-contract scanner; the rest run independently.

  2. 2

    Adjudicate

    The Oracle applies deterministic rules first, then escalates ambiguous cases to bounded LLM adjudication. Every verdict carries a confidence score and a budget audit trail — no silent dropouts.

  3. 3

    Report

    One canonical qa_report.v1.json plus SARIF and Markdown. Every finding ships a Five-Whys ladder, blast-radius matrix, and a paste-ready prompt for your IDE-AI to draft the fix.

What the output looks like

Findings, not noise.

A real finding from our self-scan. Severity says how bad; the Oracle verdict says whether it's real; the file:line says exactly where.

pipeline_contract.producer-consumer.consumer-without-producer high ◆ confirmed confidence 92
locationservices/temporal-orchestrator/src/workflows/scan_v1.ts:704
observedWorkflow awaits functional.har but the functional activity returns artifacts: [] on every code path. The API specialist downstream is silently skipped — every scan since the API specialist shipped has reported clean.
blast radiusAPI + HAR-dependent visual.diff specialists structurally unreachable. CI passes; runtime coverage is a lie.
paired withservices/temporal-orchestrator/src/activities/run-functional.ts:407
◆verdict confirmed · resolved_by deterministic · oracle/pipeline_contract@v1 See full report

Yes, this is a finding from this codebase. We dogfood the scanner on ourselves; the demo target is the demo.

Private beta opens 2026.

Concierge onboarding for the first cohort: send us your target URL and team size, we'll send back a CLI command and a link to your first report.

Request beta access

Technical readers only. We say what we caught and show the file:line.