Agents write your code, let them test it too.
UI/UX testing, automated. Describe how you want your agent to test. That's it. Watch one or more AI agents spin up real browsers and test your app like an actual user.
Provide a URL, describe your test, and run.
No SDK, no CI config, no test framework. If you can describe what a user should be able to do, you can test it.
Add your URL
Paste a URL. Add test credentials, always encrypted and never provided to the agent, plus rules for what the agent may and may never do.
Describe your test
“Add a contact and verify it appears in the list.” A sentence or two is the whole test. No selectors, no scripts, no framework.
Run
A real agent spins up a real browser in the cloud and clicks through your app like an actual user. Watch any session live, then replay every step.
Everything a QA engineer would do
SanityCheck agents run the described task in a real browser. The team gets a result and a replay of what happened.
Describe your test
Write what a user should be able to do. The agent figures out the rest and comes back with observations and a report.
A live view
Watch the agent work through your app in a real browser, in real time. Let it run in the background or take over at any time.
Forensic replay
Every action, console line, network request, and error, time-synced to video playback. See exactly what happened when.
Stress test
Launch one or more concurrent browsers at once. Part regression test, part load test, from a single trigger.
Secure credentials
Passwords are encrypted and filled by tool calls, not the model. The agent logs in without ever reading the secret.
Hard limits, every run
Allowed domains, step and time budgets, and “never do” rules, enforced by the platform on every run.
Trigger tests on every deploy
Connect GitHub, GitLab, or Vercel and mark tests “run on deploy”. The moment a deployment goes live, a team of agents is testing it.
GitHub
Install the GitHub App on a repo. When a deployment is reported successful, choose which tests launch. Nothing added to your workflow files.
GitLab
Authorize once and pick a project. Tests fire when a pipeline or deployment succeeds, filtered to the branch or environment you care about.
Vercel
Point a webhook at SanityCheck and every successful deploy, production by default, gets clicked through the moment it goes live.
“Before I ship, make sure a normal user can still do the important things in my app, without me writing test code.”
Ship like you have a QA team.
15 free agent test runs a month. No credit card, no test code, nothing to maintain.


