Agent Reliability Score v0.2

Test your AI agent in 10 minutes. No install, no API key, no account. Auto-score suggests, you confirm — so a developer gets speed, and a non-developer gets plain-language fixes.

No account · No API key · $0 · Agent answers stay in this browser

Anonymous totals only: when you score a test, an anonymous +1 count is sent (no answer, no name). See totals below. Your answers never leave this page.

Tests scored by everyone so far: — · Page loads: — (anonymous counter, can be blocked by ad-blockers, so treat as a directional total, not an exact user count).

Why this beats a manual checklist: Auto-score checks required words and values in the browser, instantly, with rules you can read in tasks.js. Why this beats DeepEval for a first test: no install, no Python, no API key, works on a phone in 10 minutes, not hours of setup. DeepEval wins for CI and automatic LLM judges — use this first, then graduate to it.

Run a test

Give this prompt to your agent

Stays in this page. Not included in exported report.
4. Score the answer (each item is 1 point)

Click Auto-score first — it suggests ticks from transparent rules. Review them, untick any you disagree with, then Score.

Developers: Download tasks.json, edit it, import your own tasks below. Export JSON for your own pipeline. No vendor lock-in.
Non-developers: You never need to touch tasks.json. Just pick a task, paste, auto-score, and read "What to fix" in the report.