Unit tests written, run, and explained

Name a function. ML.ai reads your test setup, writes the tests, runs them, and tells you plainly what passed and what broke.

Install ML.ai All use cases

New sessionAutomationsReview

Sessions

Add dueDate to tasks
Fix 401 in auth.test.ts
Tests for formatPrice

Tests for formatPriceacme/shop2m 40s

Write unit tests for formatPrice in src/utils/price.js and run them

Read price.js and package.jsonESM, test/

Wrote test/price.test.js7 tests

Ran npm testscript failed

All 7 tests pass when run directly.

npm test itself failed: its script, node --test test/, breaks on this Node version. The tests were not the problem.

Also changing package.json. Continue?

ContinueKeep to declared

Describe a task or ask a question

Mode BuildModel AutoEffortlowmediumhigh

test/price.test.jsNew file

price.test.js7 tests

1+import { test } from 'node:test'

2+import assert from 'node:assert/strict'

3+import { formatPrice } from '../src/utils/price.js'

5+test('formats cents as a dollar string with two decimals', () => {

6+ assert.equal(formatPrice(1050), '$10.50')

7+})

Who uses for tests

Developers who know the code needs tests and have not had the hour to write them.

Teams with thin coverage
Utilities that shipped with no test beside them.
Developers with a failing suite
Tests that broke after a change and need a cause.
Reviewers who want proof
A test run you can read, not a promise.

A test only counts if it ran

Generated tests are cheap to write and easy to get wrong. Here is what keeps them worth merging.

price.jspackage.jsonpasted into chat

1export function formatPrice(cents) {

2 if (!cents) return ''

3 return `$${(cents / 100).toFixed(2)}`

Your module type and your test script never made it into the chat.

What a pasted function can see

A chat window sees one pasted function

It does not know your module type, your test script, or where tests live. So it guesses.

price.js:1export function formatPrice(cents) {

package.json:1"type": "module"

price.test.js:1+import { test } from 'node:test'

The function, the project setup, one new file

It reads your setup first

ML.ai opened the function and package.json, saw an ESM project with tests in test/, and wrote the file there.

npm testscript failed

node --test test/price.test.js7 passed

What the run reported, in order

A failed run gets an explanation

npm test failed because of the project’s own script. ML.ai said so, ran the file directly, and reported 7 passing.

This turn said it would change test/price.test.js, and is now also changing package.json. Continue?

ContinueKeep to what was declared

A file it did not declare

It asks before touching other files

It said the turn would change the test file. Before editing package.json, it stopped and asked.

Writing tests by hand versus handing it off

You check the module type, copy an import, write the happy path, and run the suite last, if at all.

price.test.js, by hand

Check the module type

Write the happy-path caseDone

Think of edge casesSkipped

$npm test → not run yet

Zero and negative prices never get a test

You name the function. ML.ai reads your setup, writes the file, runs it, and reports back.

Write unit tests for formatPrice and run them

Wrote 7 tests and ran them7 passed

npm test failed on its own script. The file passes when run directly.

price.test.js7 tests · new file

1+import { test } from 'node:test'

2+import assert from 'node:assert/strict'

3+import { formatPrice } from '../src/utils/price.js'

5+test('formats cents as a dollar string with two decimals', () => {

Which agent actually does this

Build

ReadWrite testsRun

Build · Read · Write tests · Run

Writes the tests and runs them

Reads the function and project setup, writes the test file, runs it, and asks before changing anything else.

explore

ReadSearchWrite

explore · Read · Search · Write

Optional: see what is covered first

Read-only. Ask which functions already have tests before anything is written.

From one sentence to a test run

ML.ai checks how your project runs tests before it writes a line, then runs what it wrote.

Explorer

src/utils/price.js
test/price.test.js
package.json

ML.ai ML.ai

Build

general, the default agent, takes the whole task

Write unit tests for formatPrice in src/utils/price.js and run them

Mode BuildModel AutoEffortlowmediumhigh

1Name the function and the runnerOne sentence is enough. ML.ai may list the proposed work and ask which parts to do.

Explorer

src/utils/price.js
test/price.test.js
package.json

ML.ai ML.ai

Build

Write unit tests for formatPrice in src/utils/price.js and run them

Read src/utils/price.js

Read package.jsonESM, test/

Wrote test/price.test.js7 tests

Ran npm test, then the file7 passed

Describe a task or ask a question

Mode BuildModel AutoEffortlowmediumhigh

2It reads your setupIt opens the function and package.json to learn the module type and where tests go.

Explorer

src/utils/price.js
test/price.test.js+35
package.json

ML.ai ML.ai

Build

7 tests written. They record what formatPrice does today, including 0 and negative prices.

price.test.js7 tests · new file

1+import { test } from 'node:test'

2+import assert from 'node:assert/strict'

3+import { formatPrice } from '../src/utils/price.js'

5+test('formats cents as a dollar string with two decimals', () => {

6+ assert.equal(formatPrice(1050), '$10.50')

7+})

AllowAlwaysDeny

Describe a task or ask a question

Mode BuildModel AutoEffortlowmediumhigh

3It writes and runs the testsIf the run fails, you get the reason, and a broken setup is told apart from broken tests.

Explorer

src/utils/price.js
test/price.test.js+35
package.json

ML.ai ML.ai

Build

7 tests written. They record what formatPrice does today, including 0 and negative prices.

price.test.js7 tests · new file

1+import { test } from 'node:test'

2+import assert from 'node:assert/strict'

3+import { formatPrice } from '../src/utils/price.js'

5+test('formats cents as a dollar string with two decimals', () => {

6+ assert.equal(formatPrice(1050), '$10.50')

7+})

AllowAlwaysDeny

Allowed · price.test.js added, 7 passed when run directly

Describe a task or ask a question

Mode BuildModel AutoEffortlowmediumhigh

4You approve anything extraIf the turn needs a file it did not declare, ML.ai asks first. Continue, or keep to the plan.

Explorer

src/utils/price.js
test/price.test.js
package.json

ML.ai ML.ai

Build

general, the default agent, takes the whole task

Write unit tests for formatPrice in src/utils/price.js and run them

Mode BuildModel AutoEffortlowmediumhigh

Where test writing goes wrong

Three common failures, and what ML.ai did about each in a real run.

price.test.jsWith ML.ai

module typemodule, read

importsimport

locationtest/

Fits the project it is in

Tests for the wrong module system

A generated test uses require in an ESM project and fails before it tests anything.

How ML.ai ML.ai handles it

ML.ai read package.json first, saw "type": "module", and wrote import-style tests.

The runWith ML.ai

npm testfailed, cause named

price.test.js7 passed

Setup told apart from tests

A run that fails for the wrong reason

npm test goes red, and you cannot tell if the tests or the setup broke.

How ML.ai ML.ai handles it

ML.ai traced it to the script, ran the file directly (7 passed), and asked before fixing package.json.

formatPrice casesWith ML.ai

formatPrice(0)recorded

formatPrice(null)recorded

formatPrice(-250)recorded

Edge cases written down

Tests that lock in today’s behavior

A test records what the code does now, so a bug can become the expected result.

How ML.ai ML.ai handles it

The run covered 0, null and a negative price, and flagged the negative case as current behavior. Read those before merging.

Where this stops

What to expect, so nothing surprises you.

Which should I do?

AllowAlwaysDeny

It may ask first

ML.ai can list the proposed work and wait for your pick. Name what to cover to skip the round trip.

Also changing package.json

AllowAlwaysDeny

It stops at undeclared files

Editing a file it did not mention needs your OK.

runcheck

Inconclusive is not a pass

If a check has no clear result, ML.ai shows a Verification notice. Run /ml-ai:proof to see what was checked.

Root cause in session.ts

The store returns null before the session write finishes, so the middleware rejects a valid token.

session.ts+2 −1

40export async function getSession(id) {

41- const s = store.get(id)

41+ const s = await store.get(id)

42+ if (!s) throw new SessionNotFound(id)

AllowAlwaysDeny

Debug

Give ML.ai a failing test or a stack trace. It finds the actual cause and proposes a fix you approve.

See how it works

Apply edits to 3 files?

One diff across the schema, the API route and the task card.

route.tsschema.prismaTaskCard.tsx+2 −1

20export async function POST(req) {

21- const { title } = await req.json()

21+ const { title, dueDate } = await req.json()

22+ return createTask({ title, dueDate })

AllowAlwaysDeny

Build

Describe a feature in a sentence. ML.ai writes it across every file it touches, as one diff.

See how it works

FAQs

Does it run the tests it writes?

Yes. It runs them and reports the result. After a turn that changes files, ML.ai also checks the work and can make up to 2 repairs.

Will it match my test setup?

It reads package.json first, so it knows the module type and where tests go. It uses the runner you name.

What if the run fails?

You get the reason. When npm test failed on a bad script, it said so and ran the test file directly.

Can it change other files?

Only with your OK. When a turn needs a file it did not declare, ML.ai asks.

Try today, or talk to us about what is next

Install the editor agent on your own machine, or book a call to talk through your team’s workloads

Download ML.ai Book a call

Unit tests written, run, and explained

Who uses for tests

Teams with thin coverage

Developers with a failing suite

Reviewers who want proof

A test only counts if it ran

A chat window sees one pasted function

It reads your setup first

A failed run gets an explanation

It asks before touching other files

Writing tests by hand versus handing it off

By hand

With

Which agent actually does this

Writes the tests and runs them

Optional: see what is covered first

From one sentence to a test run

Where test writing goes wrong

Tests for the wrong module system

A run that fails for the wrong reason

Tests that lock in today’s behavior

Where this stops

It may ask first

It stops at undeclared files

Inconclusive is not a pass

Related use cases

Debug

Build

FAQs

Try today, or talk to us about what is next