The autonomous
sanity platform.

Give Recus a URL. It studies the page like a human tester, discovers hidden UI, crawls linked routes, and sends back a visual report — no scripts, no specs, no QA hire.

Contact sales →

Teams running sanity checks before every deploy

AcmeNorthwindHelixParcelFoundryLattice
Elements tested12,400+
Scans completed3,800+
Bugs caught890+
Tests written by humans0
Pages crawled per scan8

Product

One URL in. Full coverage out.

1# Recus auto-generated — no author required
2url: https://your-app.com/signup
3elements_found: 43
4tests_generated: 43
5coverage: 100%
6
7input[email]:
8 - empty_submit
9 - invalid_format
10 - xss_probe
$ npx recus scan https://your-app.com/signup
→ Opening page in Chromium...
Found 43 interactable elements
Generating 43 tests (0 human-authored)
✓ email input — XSS test passed
✓ submit button — click test passed
✗ phone field — accepted letters
? forgot password — no visible feedback
Report ready → recus.app/r/abc123

You ship the code.
Recus runs the sanity check.

Point it at staging, a preview deploy, or production. Recus studies each page like a human tester, discovers UI that clicks reveal, crawls linked routes, and delivers a report you can act on.

01 / 07

Page discovery

Not a script runner

Thinks like a senior QA engineer.

First look, like a person

Before touching anything, Recus scrolls the page to load lazy content, captures a full screenshot, and reads the accessibility tree — then writes a product brief describing what the page is and how it should behave.

Real QA methodology

Not a generic smoke template. Recus applies leave-one-out per required field, all-empty submit, format validation, boundary values, XSS probes, and happy paths — systematically, for every control that exists on the page.

Acts, then watches

Each test runs in a fresh browser tab with video recording. You see cursor ripples, click labels, and fill previews on every action — the same evidence a human tester would leave behind.

Judges like a reviewer

Before/after screenshots go to the vision model with the action log. Pass, fail, or warning — with plain-language details. Stuck on OTP or real credentials? Recus asks once and re-runs just that test.

Discovery & crawl

Finds what clicks reveal. Follows where links go.

Recus doesn't stop at what's visible on load. It watches the page change as it interacts, authors follow-up tests for newly exposed UI, and crawls same-origin routes until your core flows are covered.

01

Map the starting page

Recus opens your URL in Chromium, primes lazy-loaded content by scrolling top to bottom, and snapshots every interactable element from the accessibility tree — icon buttons and below-the-fold controls included.

02

Watch UI appear

During test execution, every click and navigation triggers a live tree diff. Interactable elements that weren't there on load — modals, dropdown options, forgot-password links — are recorded as discoveries with the steps that revealed them.

03

Test what opened up

Discovered elements get new follow-up tests. Each one replays the reveal path first — re-open the menu, re-click Continue — then interacts with the new control and asserts the expected behavior.

04

Crawl linked pages

Clicks that navigate to a different route are parked and queued. Recus crawls same-origin pages breadth-first — login, signup, forgot-password — giving each its own product context and test suite. External links and new tabs are logged, not lost.

Pipeline

Five steps from paste to report.

01

Study the page

1multimodal page read

Recus captures a full-page screenshot, reads the accessibility tree, and writes a product-context brief — what the page does and what should happen when someone uses it.

https://your-app.com/checkout
Run
02

Invent the tests

Autogenerated per element

Using that context, Recus generates targeted cases: required-field gaps, bad input, security probes, happy paths, and keyboard/UX checks — tailored to what's actually on screen.

discovering elements...
button[Submit] · input[email] · a[Forgot?]
running test suite (43 tests)
PASS 35 · FAIL 3 · WARN 5
03

Run in Chromium

Realclicks, not mocks

Each case gets a clean browser tab, video recording, and Playwright trace. Elements are found by accessible name, so minor DOM shuffles don't break the run.

discovering elements...
button[Submit] · input[email] · a[Forgot?]
running test suite (43 tests)
PASS 35 · FAIL 3 · WARN 5
04

Judge the outcome

Visualpass / fail / warning

Before/after screenshots and action logs go back to the model. Pass, fail, or warning — with a plain explanation and a fix suggestion when something's wrong.

Before / After

Phone field accepted "abc"

05

Discover, crawl, expand

BFSmulti-page crawl

Elements revealed by clicks get follow-up tests that replay the reveal path. New same-origin routes are queued and crawled — each page with its own context and suite. Removed UI drops off on the next scan.

input[email]3 tests pass
input[phone]3 tests fail
button[submit]2 tests pass
a[forgot]1 test warn

Workflow

Everything between your URL and the bug list.

01

Drop in a URL

No SDK install, no CI hookup, no test files. Paste staging or production and kick off a scan from the CLI or dashboard.

https://your-app.com/checkout
Run
02

Follow the agent live

Watch terminal output as Recus builds a page model, generates cases, and executes them in a visible browser window.

discovering elements...
button[Submit] · input[email] · a[Forgot?]
running test suite (43 tests)
PASS 35 · FAIL 3 · WARN 5
03

Triage with proof

Every failure includes before/after screenshots and scroll captures — so you see the bug, not just a stack trace.

Before / After

Phone field accepted "abc"

04

Answer when it asks

OTP wall, CAPTCHA, or real login credentials? Recus pauses, asks once, and re-runs only the blocked test with your input.

R
Recus Bot
🐛 3 bugs found on /checkout — View report
05

Audit the coverage

Browse every page crawled, every element discovered, and every test outcome — with video, traces, action logs, and discovered routes in one report.

input[email]3 tests pass
input[phone]3 tests fail
button[submit]2 tests pass
a[forgot]1 test warn

Sees the whole page

Vision plus the accessibility tree catches icon buttons, styled links, and controls that don't show up in a simple DOM dump — including content below the fold after a scroll pass.

Follows the flow

Progressive disclosure, multi-step forms, and UI revealed by clicks are discovered mid-run and tested with follow-up cases that replay the path to reach them.

Covers the site

Same-origin routes found during testing are crawled automatically — signup, login, checkout — each with its own context. Diff-aware reruns keep coverage current as your UI changes.

What early teams are saying

Parcel

We stopped doing the pre-demo click-through. Paste staging, read the report, fix what's red — that's the whole ritual now.

Sarah Chen / Engineering Lead, Parcel

faster pre-release checks
40+
hours back per month
0
manual test specs maintained
Northwind
43
bugs caught in month one

Zero tests authored. First report in 12 minutes.

Helix
100%
signup flow coverage

3 validation gaps found before launch.

Parcel
faster than manual QA

890+ elements tested. 1 QA hire avoided.

Foundry
28
UX warnings surfaced

Every client handoff tested before delivery.

Enterprise

Reports, history, and control at scale.

01

Any public URL

Point Recus at staging, preview branches, or production — anywhere your app is reachable over HTTPS.

02

Persistent report history

Every scan is stored with video, traces, and screenshots. Share a link instead of exporting and re-exporting.

03

Your choice of model

OpenRouter, Claude, GPT, or Grok — pick the vision model that balances cost and accuracy for your team.

04

API & CI on the way

Self-hosted agents, pipeline gates, and programmatic scans are on the roadmap. Early access users keep their rate.

From real scans

Failures Recus flagged before they shipped.

ON A FEATURE BRANCH

Phone field accepted letters at checkout

Impact: Would have blocked mobile conversions for EU customers with strict input validation expectations.

BEFORE RELEASE

Empty form submitted without validation

Impact: Signup flow allowed blank email — 100% of new accounts would have been unusable.

IN STAGING

Submit button produced no visible feedback

Impact: Users would double-click and create duplicate records. Recus flagged it as a UX warning.

ON A PREVIEW DEPLOY

XSS payload accepted in email input

Impact: Security regression introduced in a refactor — caught before merge to main.

POST-DEPLOY

Forgot password link led to 404

Impact: Broken recovery flow discovered on production URL within minutes of paste.

BEFORE DEMO

2FA prompt skipped on mobile viewport

Impact: Demo would have shown a security gap investors were specifically asking about.

Who it's for

Sanity coverage for every kind of team.

Solo founders

Ship, paste the URL, read the report. Sanity coverage without hiring QA.

Engineering teams

Replace the post-deploy click-through with an agent that runs after every push.

Product & PM

Open the report, see the screenshots, file the bug. No eng translation layer.

Agencies

Scan every client site before handoff. Look thorough without the manual pass.

Pricing

Free while we
build in public.

No credit card. No trial cap. Unlimited scans, full visual reports, video recordings, and before/after screenshots. Early users get grandfathered when paid plans launch.

What's included
  • Unlimited page scans
  • Multi-page same-origin crawl
  • Progressive element discovery
  • Full visual reports
  • Before/after screenshots
  • Video recordings & traces
  • Human-in-the-loop for credentials
  • Report history

Your next deploy deserves a sanity pass.

Run your first scan in minutes. No credit card, no test suite to maintain.

Contact sales →