GenClass finds what breaks in your web app, can hotfix it live, and hands the real fix to your AI. One command sets it up.
Join the community on DiscordThe GenClass GitHub App reads what the runtime saw while your Playwright tests ran, and comments with each race, its test and how often it came back. It never reads your source code.
GenClass runtime review
2 findings in this pull request's tests at 4f2a9c1 · 1 reproducible · 2 tests read
GenClassPOST /api/orderscheckout.spec.ts › places one order2 of 2 seeds, reproducibleGET /api/searchsearch.spec.ts › shows the results of the last query1 of 2 seedsReply @genclass-dev replay 1 for the steps, or @genclass-dev fix 1 for a brief your coding agent can work from.
A stale response, a double submit, an endpoint that keeps failing: nothing crashes, so error monitors see nothing. GenClass reports each one with its evidence, and the default mode only reports.
[GenClass] Prevented a stale write: GET /search?q=pa landed after q=paris was shown (stale; discard) [GenClass] Coalesced POST /api/orders with #212, so one order is placed [GenClass] Failing request: GET /api/inventory failed 5 times in 8 s (failing) [GenClass] Held the write to cart.total until the related in-flight operations finish, to decide again.
Over MCP, your agent gets what happened in order, reruns it with the responses reordered, and writes a test that fails on the bug.
> the search sometimes shows old results [genclass] f3 stale-write: GET /api/search?q=rea overwrote newer search.results, seen in 3 runs genclass_explain f3 → the evidence in causal order, the delivery order, how to replay genclass_replay arm → reproduced after the fix → not reproduced tests/search-f3.spec.ts written: fails on the bug, passes on the fix
What broke, on which page, and how often. Free shows one app's totals. Pro adds every problem with a fix prompt, visitor analytics, alerts and your team.
Real browsers, on held-out apps it never trained on.Every number, with its method →
01
$0no card needed
One app, its totals, and the open-source basics.
On the website. No card.
02
$15per developer, per month
Everything for up to 3 apps, per developer.
Then $15 a developer a month. Cancel any time.
03
By agreementby email
Your own classifier and our models in our cloud, on agreed terms.
By email. A person answers.
Questions about Pro, Enterprise, your own classifier, self-hosting or invoices? Email mehar@genclass.dev.
US dollars, per developer, billed monthly through Stripe. Cancel any time. Pricing questions →
Run npx genclass in your app's folder. Right after you sign in it starts Pro's 3 free days, then adds GenClass, connects the GitHub App and your coding agent, and tells you what your app needs next: the guide.
The runtime, its model and the MCP server are open source (Apache-2.0) and free, and Free on genclass.dev shows one app's totals. Pro is US$15 a developer a month for up to 3 apps, with 3 days free: pricing.
Yes. @genclass/mcp shows your agent what happened in order, replays the bug and writes a test for it. The genclass skill is one command.
The runtime needs neither: one import, and the model runs on the device. npx genclass signs you in and starts Pro's 3 free days, so your app gets a dashboard.
No. It draws nothing on your pages. What it finds goes to the browser console, your dashboard and your terminal.
The default mode only observes. GenClass introduced no bugs in 3,510 test runs across 7 frameworks, and ?genclass=off turns it off on a page.
To the browser console. The sinks option sends them to your own logging.