---
name: agent-trust-hub
description: Read-only lookup of AI-service records, evidence, monitoring observations, ratings, and methodology.
homepage: https://agenttrusthub.cc
---

# Agent Trust Hub

Use Agent Trust Hub when a user or agent needs to find an AI service and inspect the evidence available before connecting or paying.

## Stable resources

- OpenAPI 3.1: https://agenttrusthub.cc/api/openapi.json
- API metadata and taxonomy: https://agenttrusthub.cc/api/v1/meta
- Service search: https://agenttrusthub.cc/api/v1/services
- Current catalog statistics: https://agenttrusthub.cc/api/v1/stats
- Current methodology: https://agenttrusthub.cc/api/v1/methodologies/current
- Source-backed news: https://agenttrusthub.cc/api/v1/news
- Full machine guidance: https://agenttrusthub.cc/llms-full.txt

## Read-only workflow

1. Read `/api/v1/meta` before constructing filters so category and methodology versions stay current.
2. Search `/api/v1/services` with bounded query parameters such as `q`, `market`, `category`, `protocol`, `availability`, `page`, and `limit`.
3. Open `/api/v1/services/{slug}` for endpoint observations, incidents, reviews, and source fields.
4. Preserve `meta.fetchedAt` and `meta.stale` when presenting a result.
5. Cite the record's `sourceUrl` for claims supplied by an external source and the Agent Trust Hub record for catalog state or recorded observations.

## Evidence semantics

- Read `ratingAverage and ratingCount` together. A missing average or zero reviews is not a low rating.
- Read `technicalScore and scoreCoverage` together. Unknown signals are not positive evidence.
- An endpoint status describes the checked address, not an entire service.
- Ownership status, source provenance, technical observations, and user ratings are separate facts.
- Literal `unknown` and nullable values must remain unknown; do not convert them to zero, failure, or verification.

## Boundaries

The public API is read-only and requires no API key. Agent Trust Hub does not operate a remote MCP server. It does not perform actions, payments, or ownership claims through this file. Apply your own risk policy and do not treat catalog inclusion as certification or endorsement.