$19 a month, and one plan.
End-to-end tests from one sentence, on every pull request, and the tracking they keep correct. The test runs are not metered: they execute on your own CI runner and your own Anthropic key.
What the price meters is the tested pull request — a pull request your suite ran against, counted once however many times you push to it. Terminal runs and staging crons are never counted.
14-day trial at Pro limits, no card · no seats · no per-site fee · your data exports in one file any time
do not pay for this if
The unit here is the merge. At four merges a month you are paying $19 for four comments, and four comments a month is a thing you could read off a dashboard yourself. This gets good somewhere north of a merge a day.
This does not replace Plausible, PostHog, GA4 or Mixpanel. It runs beside them and reads what they and your repo already know. If consolidating vendors is the goal, this adds one.
There is no DOM recording and no video of a session. There is an event-based session inspector (pages, clicks with positions, rage-clicks, timing) and it is a different thing. If watching a recording of a user is the job, PostHog does it and we do not.
You get one. A test is a sentence, and the first run returns a verdict in about a minute against any URL you already have, with no account. What takes longer is the tracking half: a step change in an event needs a baseline, so that side has nothing useful to say until it has watched for a while.
fees, plainly
- $19 / month
- 100 tested pull requests included
- 10c each
- every tested pull request past the included ones, billed to the cent
- $6 / million events
- past the 2M fair-use ceiling, metered continuously, never a cliff
- $0.68 of AI a month
- past it, AI answers pause until the month resets; every computed report keeps working
One plan. No seats, no per-site fee, no tiers, nothing else. Annual is ten months for twelve. The repo audit and the readability check are free forever and need no account.
Already sure? Pay for the year now: $190, which is ten months for twelve. No trial to wait out, and the same 30-day money-back as everything else.
both plans get the full runner and the pull request comment, the run history, published run links you can send anyone, the exit code that fails a build only when your app fails a test, and the MCP server so your editor can run the suite.
- · one comment on every pull request, edited in place on each push: what broke, and the file most likely responsible
- · 1 year of history, so this month has something to be compared against
- · always on, so nothing waits for the instance to wake up
- · 2 isolated instances, unlimited sites on each, and sites are never the meter
- · 2M events a month of fair use, which is a ceiling rather than a price
past your included tested pull requests nothing locks: the extra ones meter at 10c each, to the cent, and every charge gets an emailed receipt
questions
We already have Playwright tests. Why would we pay for this?
Keep them. The question is not whether you can write end-to-end tests, it is whether anyone is still maintaining them in six months. This is for the flows nobody got round to covering, and for the ones that go red every time a button is renamed. When your UI changes, there is no selector to update — the agent looks at the page again and works it out.
Do test runs cost me anything here?
No, and nothing on this page meters them. The runner executes on your own machine or your own CI runner using your own ANTHROPIC_API_KEY, so the model cost is yours and it is small: a run that passes is recorded, and the recording replays afterwards with no model call at all. The comment on the pull request is posted with the GITHUB_TOKEN GitHub Actions already gives that job for free. What $19 buys is the hosted side — your project's isolated instance, the tested pull requests below, and the tracking half that keeps your instrumentation correct.
What is a tested pull request?
A pull request your suite ran against. Your CI runs the tests and posts each verdict to your project, and a distinct pull request counts once for the month — pushing to it five times is still one, because the unit is the change we stood behind, not how often you pushed. Runs with no pull request behind them are never counted: someone at a terminal, a cron against staging, a replay on your laptop. Those are free because metering them would teach you not to run tests.
How much does it cost?
Pro is $19 a month and includes 100 tested pull requests, then 10c each. A tested pull request counts once however many times you push to it, and local runs are never counted, so you pay for changes stood behind rather than for data stored. The repo audit and the readability check stay free and need no account.
Is anything free?
Two things, and neither asks for an account: npx smolanalytics audit reads the repo you are standing in and names the user actions nothing is measuring, with the file and the line and no network call, and the AI readability check tells you whether an AI crawler can read a given page. What there is not is a perpetual free cloud tier. A hosted instance costs us a machine every month, and a plan we lose money on is one we would have to withdraw later, so the 14-day trial is the free version of the hosted product.
What does the trial not include?
Model calls. Three features spend a real model call each time they run, and a $0 plan funds none of them: the model-routed ask bar, AI-visibility sampling, and the runner that opens an instrumentation pull request. The trial's instance also sleeps when it is idle and keeps 14 days of history rather than a year. Everything computed works: the run history, the reports, the CI gate, exports, and the MCP server in your editor.
What happens if I go past the included events?
Nothing breaks and nothing samples. Events are a fair-use ceiling rather than the price, set at 2M a month. Past it we keep ingesting and meter the excess at $6 per million, continuously rather than in whole blocks, so one event over costs a fraction of a cent instead of a whole $6. You get an email at 80% and again at 100% before any of that, and reads never lock.
Can the AI run up my bill?
No. The AI allowance is inside the price you already pay. When a month's allowance is spent, AI answers pause until it resets rather than quietly billing you, and every computed report keeps working while they are paused, because the reports are computed rather than generated. That is the whole of the AI line on your invoice: there isn't one.
Is this cheaper than PostHog or Mixpanel?
It is not a replacement for them, so that is the wrong comparison and we would rather say so than win it. They collect your events; this runs your end-to-end tests and keeps the tracking calls inside their SDK correct. You keep whatever you collect with today, and the price is an addition to that bill rather than a substitution for it. If what you actually want is a cheaper collector, you want Plausible or Umami, not us.
What happens if I stop paying, or never pick a plan?
You keep working through a grace period, and nothing is deleted quietly. Machines stop after 7 days with the data intact, a final export notice goes out by email on day 30, and removal is not possible before day 37 or sooner than 7 days after that notice. Everything you collected exports in one file at any point along that timeline.
Not sure what a test looks like? how it works walks one from the sentence to the comment on the pull request.