500+ AI agents exist, but no way to know which ones actually work. Benchmarks evaluate LLMs, not agents. Two agents on the same GPT-4o can have wildly different reliability. Legit evaluates agents, not models. 36 tasks, 3 AI judges (Claude + GPT-4o + Gemini), one trust score. Three commands. Zero cost. Five minutes. Open source, Apache 2.0.