AgentTrust Dashboard

Pre-payment quality verification for AI agents, MCP servers, and skills. Multi-judge consensus scoring across 6 dimensions.

Total Evaluations

Average Score

Pass Rate

Expert Agents

Avg Eval Time

Tools Tested

Tier Distribution

Evaluation Standards

Multi-Judge Consensus

2-3 parallel LLM judges with agreement threshold

6-Axis Scoring

Accuracy, Safety, Reliability, Latency, Process, Schema

Adversarial Probes

Injection, extraction, PII, hallucination, overflow

AQVC Attestation

Ed25519-signed JWT with W3C VC compatibility (AgentTrust Verified)

Anti-Gaming

Question paraphrasing, canaries, production correlation

Recent Evaluations

How AgentTrust Works

Connect

Paste any MCP server URL. We auto-detect SSE or Streamable HTTP transport.

Discover

Automatically list all tools, validate manifest schema, check documentation quality.

Evaluate

Run functional tests per tool. Multiple LLM judges score responses independently.

Attest

Generate 6-axis quality score, tier badge, and Ed25519-signed AQVC attestation (AgentTrust Verified).

AgentTrust v1.0 — Evaluation Version v1.0