Verdict-first reporting
Every report opens with an evidence-supported release assessment, including Inconclusive when workflow coverage is insufficient.
FeaturesBuilt for Android release review.
Built for Android release review.
Not generic QA.
Caraxe keeps the scope bounded: evidence first, regression memory across builds, and a release decision teams can act on.
Accessibility auditing
Accessibility defects are product defects. Caraxe shows the control and the evidence.
Caraxe audits Android interfaces for missing accessibility labels, unlabeled icon controls, and touch targets smaller than Android’s recommended 48 dp interactive area. Findings retain the screen evidence and the measured control geometry, so a team can reproduce and prioritise the issue instead of receiving a generic accessibility score.
- Verified classes: controls a screen reader cannot announce, icon-only controls without a meaningful label, and touch targets below the 48 dp guidance.
- Each finding names the rule that produced it. Android’s 48 dp recommendation is deliberately stricter than the WCAG 2.2 Level AA minimum in many mobile contexts, so the report says which one it measured against.
- In the sealed benchmark, an accessibility catch counts only when the persisted finding exactly matches the hidden defect class after the relevant screen was mechanically reached.
- Accessibility findings published on real apps were re-measured by an independent reviewer on a physical device before appearing on this site.
This is guideline-aligned accessibility auditing. It helps teams find and document accessibility risk earlier. It is not an accessibility certification, a legal opinion, or a guarantee of WCAG, ADA, Section 508, EN 301 549 or any other regulatory conformance, and it does not replace human and assistive-technology testing.
Build-over-build memory
Findings stay threaded across builds so teams can see what is new, recurring, resurfaced, escalated, or resolved.
Reproducible evidence
Findings carry a concise evidence bundle, with trace-backed reproduction context whenever the recorded action path supports it.
Impact taxonomy
Customer-readable issue buckets map directly to engineering ownership and release risk.
Systemic insight detection
Repeated patterns across screens are grouped as one operational insight instead of noisy duplicate findings.
Trust by design
Private artifacts, pixel redaction, disclosed AI processing, and findings grounded to runtime evidence.
Also included
The details that make the report usable.
Screenshot diff slider
Compare current and baseline screens without opening a second tool.
Zero-setup device
The device path is implemented; production capacity and operating recovery checks remain launch gates.
API and CI gating
Integration surfaces exist in code; automated release gating stays closed until the core is certified.
Team workspaces
Workspace and role surfaces are implemented; public multi-team operation is not yet open.
Calibration loop
Feedback capture is implemented; measurable cross-app calibration value remains a certification question.
Replay reviews
Replay behavior is implemented; speed and cross-app reliability certification remain open.
Help validate it on real release work.
Private validation is in progress. We are accepting a small number of Android teams whose apps fit the current supported scope; public access waits for multi-app certification.