Evidence for every run

When a test breaks, see exactly what broke and why.

Smoketest monitors your critical tests and hands you the evidence on every run: a recording, a transcript, a result with reasoning, a screenshot at the failing step, and a Playwright trace.

Recordings kept for 90 days. Your first run in about two minutes.

smoketest.sh/dashboard
Board
Tickets currently moving through QA
Development1
Jira · APP-917
Invite member from workspace settings
App QANot tested
Testing2
Linear · WEB-284
Apply annual discount at checkout
Checkout QARunning
Linear · AUTH-126
Recover account with magic link
Authentication QANeeds attention
Passed1
Jira · BILL-442
Show updated invoice after seat change
Billing QAPassed
The problem

A failed test shouldn't start a scavenger hunt.

Just a red X

A test that only says “failed” sends you hunting through CI logs.

Can't reproduce

The bug won't repro locally and you're guessing what the user did.

No shared context

You can't hand a teammate proof of what actually broke and where.

Black-box AI

AI testing that won't show its work is impossible to trust.

What every run hands you

The whole story, not just a red X.

A test that only says “failed” sends you hunting. Here's what a Smoketest run gives you instead, every time.

signup-test.webm
00:00↑ 00:06 · sign-up00:07
step log
Opened /pricing
Clicked “Start free”
Filled email · password
Submitted sign-up form
Expected dashboard · got 500 error
Session recording
Screenshot at failure
Step-by-step transcript
Downloadable Playwright trace
The artifacts

Six pieces of proof for every run.

Session recording

Video of the real browser for the whole run, scrubbed to the moment it broke. Kept 90 days.

Step-by-step transcript

Every action in order, with a marker on the failing step and what it expected versus found.

Result with reasoning

A pass or fail result plus the agent's explanation of what it checked and why.

Screenshot at the failure

A frame captured the moment the journey broke, so you see the error state at a glance.

Playwright trace

A standard trace an engineer can open in the viewer to inspect the run in detail.

Shareable run link

Hand the same recording, transcript, and result straight to whoever can fix it.

From test to proof

One sentence in. A full report out.

you write
Start on the pricing page.
Choose the Pro plan.
Create a new account with the test email.
Complete checkout using the test card.
Confirm the app shows the billing success screen.
when it fails, you can
Watch the recording cued to the step that failed
Read the transcript to see exactly what the agent expected
Open the screenshot of the error state
Download the Playwright trace to dig deeper
Share the run link with whoever needs to fix it
How it works

Jira or Linear starts the Test.

Keep the workflow where QA already works. Smoketest adds execution, evidence, and routing without becoming another board.

workspace · connections
Lhealthy
Linear
Acme workspace
Jhealthy
Jira
acme.atlassian.net
01

Connect Jira or Linear

A workspace owner authorizes Jira or Linear once. Smoketest keeps Connection identity, permissions, token health, and webhook health visible.

new test source
Listen to
Checkout board · QA
Passed
Ready to release
Failed
Development
Run at
staging.acme.com · Desktop
Permissions and direct transitions validated
02

Define the QA handoff

Choose the team or board, exact QA status, direct Passed and Development statuses, then set the URL, Environment, device, and access settings.

test request · attempt 1
WEB-284 enters QA
Ticket + linked PR frozen
Generated Test created
Browser Run in progress
03

Let the ticket start testing

When a ticket enters QA, Smoketest freezes current context, enriches it from an eligible ticket-linked pull request, generates one Test, and runs it automatically.

result · synchronized
PASSEDWEB-2841m 42s
Attempt comment updated with evidence
Authenticated recording and transcript linked
QAReady to release
04

Return the Result safely

Smoketest posts evidence to one evolving Attempt comment. Passed moves forward, Failed returns to Development, and No Result stays in QA. Newer human changes always win.

Pricing

Evidence included with every run.

The recording, transcript, result, and trace come standard: no add-on, no separate tier.

14-day free trial · 50 credits · no card
Failure and recovery alerts to email, Slack, or Discord
Recordings, transcripts & traces kept for 90 days
FAQ

Questions about run evidence

Open the run. Smoketest shows a pass or fail result, the reason it failed, a screenshot at the failing step, the full session recording, and a step-by-step transcript. You can see exactly what the agent did and where the journey broke, without reproducing it by hand.

Every run produces a result with reasoning, a session recording of the browser, a step-by-step transcript, a screenshot at the point of failure, and a downloadable Playwright trace. You can also copy a shareable link to the run.

Yes. Each run is recorded as a video of the real browser so you can watch the whole journey play back, scrubbed to the moment something went wrong.

Yes. The transcript lists every step the agent took, with a clear marker on the one that failed and what it expected to see versus what it actually found. The recording is cued to that step.

Smoketest shows its reasoning. Each result comes with the agent's explanation of what it was checking and why it decided the test passed or failed, so you're never trusting a black box.

Yes. Every run includes a downloadable Playwright trace, so an engineer can open it in the Playwright trace viewer and inspect the run in detail.

Run recordings are retained for 90 days, so you can go back and review what happened on a past run when you need to.

Yes. Copy a link to the run and send it to a teammate. They see the same recording, transcript, result, and artifacts, useful for handing a regression to whoever can fix it.

Usually, yes. The recording shows what the user would have seen, the transcript shows each step and the failed expectation, the screenshot captures the failure, and the trace lets an engineer dig deeper, often enough to find the cause without re-running it locally.

Every run. Passing runs save the recording, transcript, and result too, so you can confirm a test really worked and review it later, not just inspect the failures.

Stop guessing why a test failed.

Run one test and watch the recording, transcript, and result come back. See exactly what broke, without reproducing a thing.

Read the docs