Benchmark how AI agents choose your MCP tools.

Paste a public MCP URL. The free scan reads tools/list, generates description-derived routing cases, and checks for obvious selection collisions. It never executes your MCP tools.

Public URLs only. Local/private network targets are blocked. The free score is generated-readiness evidence from the same tool descriptions — not untouched natural-language or real-model routing accuracy, and not a claim about global model discovery ranking.

ARD manifest · 10-MCP benchmark · llms.txt