MonkeyTest agents use your site like a real user to find bugs, report them, and verify them once fixed.
Free forever · BYO LLM or local · No credit card on cloud
Built for
The problem
Coding agents and vibe tools write apps faster than anyone can write tests. Playwright Agents help when you already know the journey. They don’t tell you what you never scripted.
Contact/API fails; the page looks fine.
Checkout adjacent paths, empty states, dead ends.
Until a messy user (or your customer) clicks sideways.
Selector healed; the bug may still be live.
Most gaps aren’t in your unit tests. They’re in how a messy user actually behaves.
That’s why AI-built software needs a ship-check — not another suite to maintain.
How it works
Paste a URL. No suite to author first.
Point the agent at a URL. It explores in a real browser — clicks, forms, paths scripts miss — and files bugs with screenshots.
You patch the product (or your coding agent does). Keep the same plan.json.
monkeytest rerun + diff — same plan, fingerprint diff, CI exit codes. Know the bug is gone.
Two ways to use it
The open-source CLI and library are the main path. The free cloud needs no install and no API key if you want to watch a run first.
Open-source CLI and library. BYO LLM or Ollama. Unlimited local runs. CI-ready.
No install, no API key. Watch a Scout run. 2 attempts/day. No paid plans — self-host for unlimited.
Same engine either way. Cloud is the demo; CLI is how you keep using it. It's free — no paid plans.
Open source
The engine is free and open source. Bring your own LLM, or run a local model. The live demo only lets people watch it work.
$ npm install -g @monkeytestai/monkeytest-core
$ monkeytest run https://example.comAGPL-3.0. Free forever. You pay your own LLM provider, or nothing with a local model.
Open an issue or read CONTRIBUTING.
Local models run through Ollama. Any OpenAI-compatible endpoint works too.
Where we sit
Browserbase and browser-use give agents a browser. We answer: what broke, and is it fixed?
| Question | monkeytest | Playwright Agents / MCP | Browser infra (Browserbase, browser-use) | Suite tools (Momentic, etc.) |
|---|---|---|---|---|
| Job | Exploratory ship-check + prove loop | Help you author/heal tests you define | Sessions / agent SDK | Maintain journeys you wrote |
| Needs a suite first? | No — paste URL | Usually yes (intent/seed) | N/A (not a QA product) | Yes |
| Prove the fix | First-class rerun + diff | Re-run your tests | No | Heal/re-run authored tests |
| Cost shape | Free OSS + BYO LLM | Free (Microsoft) | $/browser-hour | Credits / seats |
| Best when | Preview URL, unknown unknowns | You already know the path | Building any web agent | Regression at scale |
We don’t compete with Browserbase — we can sit on top of it later. We compete with “we installed Playwright Agents and still don’t know what we didn’t write tests for.”
Open source forever. Cloud demo when you want to watch first.