Launch HN: TesterArmy (YC P26) – Agents that test web and mobile apps
tester.army okwasniewski
A YC startup called TesterArmy launched agents that test your web and mobile apps automatically, before and after deployment. You describe tests in plain English instead of writing scripts, and it already caught checkout and login bugs for 30+ teams.
Testing has always been the part of shipping software that nobody actually enjoys. You write the feature, you ship it fast thanks to AI coding tools, and then you hit the wall: someone still has to click through the checkout flow forty times to make sure nothing broke. TesterArmy, launched on Hacker News by founders Oskar, Szymon, and Piotr, wants to hand that grind to an agent instead of a person.
The pitch is simple on paper. You describe a test case in natural language, point a CLI at your app, and TesterArmy's agent runs it end-to-end, on a schedule or triggered by a GitHub push. No selectors to maintain, no brittle scripts that break the moment a button moves three pixels to the left. When something fails, the team gets pinged in Slack or Discord instead of finding out from an angry customer.
What makes the launch post interesting isn't the framing, it's the receipts. In a few months TesterArmy went from zero to more than 30 teams running tests daily, and the founders shared specific bugs their agents caught before they reached production: a timezone bug buried in a complex booking dashboard, an agent-orchestration regression that froze a sandboxed environment, a checkout flow silently miscounting order totals, and a broken tool-calling issue that would have locked users out of their own data in an AI chat feature. These aren't hypothetical demo bugs. They're the kind of quiet, ugly regressions that hide in dashboards and chat flows precisely because they're tedious for a human tester to reproduce.
The company is also leaning hard into the idea that its own product is agent-native, meaning your coding agent, not just your engineers, can define and run tests through TesterArmy's platform. That's a deliberate bet that the near future of QA looks less like humans writing Playwright scripts and more like agents talking to agents, with a person mostly checking Slack for alerts. Whether that holds up at scale, especially for messier legacy apps with weird auth flows, is exactly the kind of thing the 69-comment HN thread is arguing about right now.
My take
Testing-as-a-bottleneck is a real problem, and letting an agent own the tedious parts is an obvious next step now that coding agents already write half the code anyway. The bugs they listed, timezone quirks, checkout miscounts, are the boring-but-costly kind that actually move revenue, which is a better pitch than another flashy demo. The real test isn't the launch thread, though, it's whether this still works when your app has five years of legacy auth logic and nobody remembers why the staging database is broken.”
Read more about this at: tester.army
Related stories
Launch HN: Intuned (YC S22) – Build and run reliable browser automations as code
intunedhq.com · 1 month ago ·
31
Launch HN: Hoplite (YC S26) – Effortlessly deploy cloud coding agents
hoplite.sh · 3 days ago ·
50
Launch HN: Coasty (YC S26) – An API for computer-use agents
coasty.ai · 3 weeks ago ·
28