letsrunit – Testing for agentic workflows

Stop guessing. Start testing.

AI agents move fast. Things break.
Letsrunit shows what and why.

Built for AI coding agents

Tests as readable steps

AI writes code you don’t read. Tests in code don’t help, you won’t read those either.

Letsrunit uses clear, step-by-step instructions. Each scenario shows what a user does and what should happen, so you can quickly verify what your AI built.

Scenario: User logs in

When I set field "Email" to "user@example.com"

And I set field "Password" to "password123"

And I click button "Login"

Then I'm on page "/dashboard"

And I see that the page contains text "Welcome"

The test expresses intent. Playwright runs underneath, but the logic is clear without reading code.

Replace ad-hoc checks with reliable tests

You click through your app to see if it still works. That works at first, but as your app grows, there are too many paths to check.

Letsrunit lets an AI agent check those flows for you. Each check is saved as a test, so instead of a one-off “ok”, you build up a test suite as you go. That suite runs on every change, so regressions are caught automatically.

Search and view results

Add item and checkout

Edit profile settings

Reset password

Failure explanations

When a test fails, the page from the failing run is analyzed against the page from a previous successful run. This makes it possible to explain why the test failed in plain language and how to solve it.

Update required

A new checkbox labeled "I agree to the terms & conditions" has been added to the registration form. This checkbox is required and must be checked before the form can be submitted.

Update your test to check this box before clicking "Register".

Possible regression

The "Proceed to payment" button was removed from the cart page. With this button gone, there's no clear way for users to continue to payment from the cart.

This appears to be a regression. The button should be restored, or the checkout flow has fundamentally changed.

Run tests in the terminal

Generated tests run with Cucumber. Install the package, use the step definitions we provide, and run the tests like any other Cucumber suite.

$ npx cucumber-js

......F-.

Failures:

1) Scenario: Visitor sees the homepage greeting # features/homepage.feature:3

Last passed: commit d498a33, 1 commit ago

✔ Before

✔ Given I'm on the homepage

✖ Then the page contains text "Hello world"

URL: /

Locator: text=/Hello world/i

Error: element(s) not found

- And I should be on page "/"

✔ After

2 scenarios (1 failed, 1 passed)

5 steps (1 failed, 1 skipped, 3 passed)

0m07.028s (executing steps: 0m06.698s)

$ npx letsrunit explain

features/homepage.feature :: Visitor sees the homepage greeting

✔ Given I'm on the homepage

✖ Then the page contains text "Hello world"

- Then I should be on page "/"

Possible regression

The page heading text was changed from "Hello world" to "Hllo world", so the expected greeting is missing.

💡 Fix the homepage text to read exactly "Hello world".

The tests integrate into existing CI. This is not a new test ecosystem. It's standard Gherkin running on standard tooling.

Standard tooling

Cucumber executes the scenarios. No custom runners or proprietary apps.

CI ready

Run in GitHub Actions, Jenkins, or any CI. The tests are files in your repo.

Run tracking

Each test run is stored locally with its artifacts, including HTML and screenshots.