Start free. Upgrade when you need unlimited testing.
Never. Your Anthropic API key lives only in your browser session. We have no backend that touches it.
We test against claude-3-5-sonnet-20241022 (the 4.6 era model) and claude-sonnet-4-5 (the 4.7 model). These are the two versions where regressions have been most reported.
One test run = clicking the Run button once. It calls both models with all your test cases. Free users get 3 runs total.
Yes — that's the whole point. Add any prompt you use in production, define what the output should contain, and see which model version passes.
The dashboard will show 0% for both. This usually means your expected output string needs adjusting, or the prompt needs work regardless of model version.