AI testing platform

Tests written in plain English. Run in CI. Healed when the UI changes. Triaged when they fail.

An AI testing platform for teams that ship faster than they can write and maintain test code.

tests/checkout.test.yaml

1name: Checkout with a saved card2url: https://app.example.com3tags: [checkout, smoke]4env:5  EMAIL: ${TEST_EMAIL}6  PASSWORD: ${TEST_PASSWORD}7steps:8  - Log in with EMAIL and PASSWORD9  - Open the cart10  - Click "Checkout"11  - Choose the saved card ending in 424212  - Click "Place order"13  - assert: The order confirmation page shows an order number14  - api:15      get: /api/orders/latest16      expect:17        status: 20018        body.status: confirmed

Product preview

Numbers

1,240

tests created

workspace count on the given date, as of 2026-10-15

86,500

test runs executed

cloud and CLI runs reported to the platform, as of 2026-10-15

312

bugs caught

failures triaged as app-bug and confirmed by the team, as of 2026-10-15

91%

healed steps accepted without edits

2,140 healed steps reviewed between 2026-09-01 and 2026-10-15, as of 2026-10-15

4,870

pull requests verified

PR checks reported by the CI integration, as of 2026-10-15

11 minutes

median time from spec to first green run

380 new tests created between 2026-09-01 and 2026-10-15, as of 2026-10-15

6 minutes

median cloud run duration for a 50-test smoke suite

1,900 cloud runs, 4 parallel workers, as of 2026-10-15

38%

failures classified as test-issue rather than app-bug

2,600 triaged failures between 2026-09-01 and 2026-10-15, as of 2026-10-15

87%

triage classification confirmed by the team

1,150 triage cards with a human verdict, as of 2026-10-15

healed steps that turned out to be behaviour changes

2,140 healed steps reviewed, as of 2026-10-15

40 seconds

median time to triage card after a failed run

all failed cloud runs in the period, as of 2026-10-15

What teams say

We moved the checkout suite to plain-English specs in a week. Since then a redesign of the cart page produced healed steps instead of a red pipeline.
Head of Engineering, e-commerce, 40 engineers

Most of our code is written by agents now. The triage card is the first thing a reviewer reads on a failed check.
Example Name, CTO, Example SaaS

Our QA engineer used to spend two days after every release fixing selectors. Now she reviews a list of healed steps in twenty minutes and spends the rest on exploratory testing.
QA Lead, fintech, 25 engineers

The specs are readable by our product manager. She wrote the acceptance steps for the onboarding flow herself and they ran without an engineer touching them.
Example Name, VP Engineering, Example Marketplace

We hand the tests/ folder to the client at the end of the project. Plain YAML, no framework to learn, and it keeps running in their CI after we leave.
Founder, product studio, 12 people

A flaky test used to mean a retry button. A triage card that says environment with the network log attached means we fix the staging DNS instead.
Platform Engineer, healthcare SaaS

Cursor writes the feature, the agent writes the spec in the same PR, and the check tells the reviewer whether the flow still works. That is the whole review process for small changes now.
Example Name, Head of Engineering, Example Internal Tools

Case studies

Checkout suite migration at an e-commerce team

64 tests migrated, 3 app bugs found in the first week

Cutting post-release selector fixes at a fintech team

2 days per release to 20 minutes of healed-step review

Agent-written specs in an internal-tools team

140 specs written by coding agents in 6 weeks, 9 app bugs found before merge

Code is shipped faster than tests are written

In the Stack Overflow developer survey, 84% of respondents use or plan to use AI tools in their development process [Stack Overflow], and 51% of professional developers use them daily [Stack Overflow]. Code is written faster than a person can check it by hand.

End-to-end tests written in code need an engineer to write them and, more often, to fix them. Tests also fail without a code change: at Google, almost 16% of tests showed some level of flakiness [Google Testing Blog]. When time is short, test maintenance is the work that gets cut, and a suite nobody maintains stops being read.

Our reasoning: if a test is plain English, more people can write and read it; if cosmetic UI changes are healed and behaviour changes fail, maintenance drops to decisions a person needs to make.

What the platform does

Tests in plain English

A test is a list of steps in plain English in a YAML file; the platform resolves each step against the live page at run time.

Runs in CI

One CLI command runs a suite in CI and reports a pass, fail or healed status per test as a check on the pull request.

Heals locator and layout changes

When a step no longer resolves because an element moved, was renamed or restyled, the platform re-resolves it, marks the test as healed and records the change for review.

Triages failures

A failed test gets a triage card with the failing step, the page state, the diff from the last passing run and a classification: app bug, test issue, environment.

Parallel hosted runs

Runs tests in parallel on hosted browsers in the selected region with traces kept for the plan's retention window.

All capabilities

Write, run, triage and heal

Step 1

Write the test in plain English

A test is a YAML file with a name, a start URL and steps in plain English. The platform resolves each step against the live page at run time.

tests/login.test.yaml

1name: Log in with email2url: https://staging.example.com3steps:4  - Click "Log in"5  - Type EMAIL into the email field6  - Type PASSWORD into the password field7  - Click "Continue"8  - assert: The dashboard shows the account name

Step 2
Run it locally or in CI
One command runs the suite locally in Chromium, or in hosted browsers with --cloud. In CI it reports pass, fail or healed per test as a check on the pull request.
terminal
```
$npx aitp run --tag smoke --cloudPassCheckout with a saved cardHealedLog in with emailexit code 2: healed steps pending review
```
Step 3
Triage failures, review heals
A failed test gets a triage card with the failing step, the page state, the diff from the last passing run and a proposed classification. Healed steps wait for review.
Product preview

How it works in detail

Learning your product

Beta Builds a map of flows from session recordings, product docs and the codebase to propose tests and resolve ambiguous steps.

Does not read production user data; sources are chosen and connected by the team.

Learning your product Coverage gaps

Product preview

Use cases

SaaS products

onboarding, billing, permissions, settings

E-commerce

search, cart, checkout, payments, account

Fintech

KYC, transfers, statements, limits

Healthcare

scheduling, intake forms, portals, consent

Marketplaces

listings, messaging, payouts, disputes

Internal tools

admin panels, approvals, reports

Security

Cloud runs in EU or US region chosen per projectVerified

Traces and snapshots encrypted at rest and in transitVerified

Test credentials stored encrypted, never shown in reportsVerified

Retention window per plan, deletion on requestVerified

Security, including items in progress

Bugs the platform caught

Saved card selection reset after a coupon was applied 2026-10-02
Order confirmation page showed the previous order number on a slow network 2026-10-09
Password reset link opened the login page instead of the reset form when the email contained a plus sign 2026-10-11
Invoice total ignored a percentage discount after currency switch 2026-10-13
Admin role could not see the export button after the permissions refactor 2026-10-16

Questions

What is an AI testing platform?

A tool that uses AI to write, run or maintain software tests. AI Testing Platform runs end-to-end tests written in plain English, heals locator and layout changes and proposes a classification for each failure.

Does it work with my CI?

Yes, with GitHub Actions, GitLab CI and CircleCI. Jenkins is in beta. The CI step runs npx aitp run.

Can I leave and keep my tests?

Yes. Specs are YAML files in your repository and can be exported to Playwright tests in TypeScript.

Request early access

Early access: prices are the planned list prices. Billing starts when the platform enters beta; early-access use is not charged. We reply within 24 hours on business days.

Request early access See how it works