Agent Trust Lab

Independent evidence for how reliably AI agents select MCP tools, with frozen benchmark results, preserved semantic misses, multi-batch reliability evidence, and independent MCP evaluation.

Documentation

MCP provider evidence and decision workspace

Choose with evidence.Review before action.

Discover MCP providers, inspect evidence by capability and version, and review task-specific options before use.

Evidence from official sources Clear evidence coverage grade Decision guidance for AI agents Transparent and up to date

Featured MCP Providers

Loading the imported catalog… · source metadata and data quality grades.

Browse all listings →

Loading locally imported Registry listings…

Registry entries support discovery only. Their grades describe the official listing data; endpoints are not checked and the records are not Trust Live recommendations.

Trust Live provider evidence

Loading providers… with non-synthetic, evidence-backed records.

Review a task →

Loading Trust Live provider evidence…

Evidence grades describe capability-specific coverage only. They do not certify safety, rank providers, or authorize an action.

Evaluate an MCP metadata endpoint

Inspect public discovery metadata for a bounded readiness report. This does not call the provider's business tools. The free scan generates cases from tool descriptions; it does not measure untouched natural-language or real-model routing accuracy.

Recent Registry Changes

Browse Registry →

Checking locally validated snapshot history…

Changes are compared only between locally validated Registry snapshots.