Sign in Start testing

01 / Overview

Continuous testing for production softwareRef. QA-01

Software quality. On autopilot.

An AI QA engineer that continuously tests your web apps, APIs, and critical workflows — catching regressions and explaining failures before they reach your users.

Start testing Explore platform
Execution · livethis page / your browser
Result
QUEUED0 of 6
  • WAITHeading “Software quality. On autopilot.” is visible—
  • WAITDocument title is set—
  • WAITButton “Start testing” is visible and enabled—
  • WAITLink “Explore platform” resolves—
  • WAITNo horizontal overflow at this width—
  • WAITNo script errors since load—
Fig. 01 — A real run against this page, in your browser. Watch the dashed boxes: that’s what each step is checking.
T-04 · Checkout CTAFailed
staging.demo-shop.test/cart
title ✓
button “Checkout” · not found
assert getByRole('button', { name: 'Checkout' }).toBeVisible() ✕ not visible within 5000 ms network GET /api/cart/summary → 500 console 0 errors · screenshot saved

02 / Problem

Why tests stop getting written§ 02.0
Position

We build for the team that knows what should be tested and never has the hours to keep it tested.

A test suite is only as good as the last time someone maintained it. Softsurance keeps testing after people stop.

01

Coverage decays quietly.

New pages ship without tests. Old tests break on a renamed selector and get skipped. Nobody decides to stop testing — it simply stops.

02

A red build is not an answer.

Knowing that something failed is the easy part. The hours go into finding the screenshot, the console line and the request that returned 500.

03

Users find it first.

Without a recurring check against staging and production, the first report of a broken checkout is a support ticket.

03 / Products

What Softsurance tests5 modules

Five ways in. One evidence record.

Browser testing ships first. API, database and end-to-end workflow testing follow on the same evidence record.

01

Frontend testing

Real browsers, real pages, recorded evidence.

Release 102

API testing

OpenAPI import, status and schema assertions.

Preview03

Database testing

Read-only schema and integrity checks.

Roadmap04

End-to-end workflows

Structured journeys with approval gates.

Roadmap05

AI investigation

Evidence-grounded failure reports.

Preview

04 / Console

The execution recordEvidence view

One failure, and everything you need to read it.

Assertions, screenshots, console output and failed requests stay together, so whoever is debugging starts with facts instead of a search.

demo-shop / stagingSuite · Storefront smokeScheduled · daily 06:00 ET● Failed
T-01 Homepagepass
T-02 Titlepass
T-03 Navigationpass
T-04 Checkout CTAfail
T-05 Cart consolepass
T-06 Loginblocked
expected · button “Checkout”
assert getByRole('button', { name: 'Checkout' }).toBeVisible() ✕ element not found within 5000 ms console 0 errors network GET /api/cart/summary → 500
Timeline
Step 4 · +5.0sAssertion failed. Screenshot and DOM snapshot captured.
Step 4 · +3.1s/api/cart/summary returned 500 before render.
HistoryPassed in the previous scheduled run.

05 / Method

From URL to evidenceHuman-approved

What happens between your URL and a result.

Five stages. Nothing executes until a person has approved it, and nothing leaves the scope you set.

● 01

Discover

Playwright loads the target, follows same-origin links within depth and page limits, and records pages, forms and errors. Nothing is submitted.

Gate · scope
○ 02

Plan

Sanitised findings become proposed tests in a validated schema. Without an AI provider, deterministic templates are used.

Gate · schema
○ 03

Approve

You edit, approve, reject or exclude each proposal. Anything that could change data needs explicit approval.

Gate · human
○ 04

Execute

Approved tests run in an isolated worker as a restricted action set. No AI-written code is ever executed.

Gate · policy
○ 05

Investigate

Failures are analysed against the evidence actually captured, with facts and hypotheses kept apart.

Gate · evidence

06 / Continuous

Deterministic schedulingTimezone-aware

Set the cadence once. It keeps running.

Schedules live in the database and run through a real task queue — never a browser timer, and never an AI deciding whether to run.

Shortest interval15min

Also hourly, daily, weekly, once, or a custom cron expression.

Overlapping runs1max

Duplicate-run protection when a single active execution is required.

Skipped runsLogged

Every skip records why: limit reached, overlap, or schedule paused.

Silent retries0

Retries are bounded and recorded. Flaky tests are shown, not hidden.

Start

Your first runStaging URL required

Add a staging URL. Read your first real result today.

Start testing Read the docs