Developers

Agent Trust Hub data for applications and AI agents

The public read-only API exposes the catalog, current coverage, observed incidents, and active methodology while keeping provenance, reviews, and technical signals separate.

A versioned public contract

The API foundation lives under /api/v1. Responses include data and snapshot metadata: fetch time, stale-cache state, and API version.

Connect an agent with one request

Public reads need no API key. An agent can filter by market, category, protocol, access method, and endpoint state, then apply its own policy to ratingAverage, ratingCount, technicalScore, scoreCoverage, sourceUrl, and lastCheckedAt as separate fields.

Zero reviews is not a low rating, and an endpoint state does not describe the entire service. Use meta.fetchedAt and meta.stale to account for snapshot freshness.

GET https://agenttrusthub.cc/api/v1/services?market=russia&category=automation&availability=healthy&limit=20&sort=score

Evidence instead of a ready-made verdict

The API keeps record provenance, user rating, technical index, and endpoint state in separate fields. Unknown information stays unknown, and an endpoint state is not presented as the state of the whole service.

An application or agent can therefore define its own risk tolerance and show the facts behind a choice.

OpenAPI and updates

The contract is described in OpenAPI 3.1. Read-only responses use short-lived caching and support browser clients. The JSON API, RSS, and website read one published feed. In live mode it shows entries from allowlisted official sources; when that mode is unavailable, the site clearly labels its reviewed editorial collection.

What this release does not include

The current public foundation has no API keys, remote MCP, signed snapshots, wallets, or paid checks. A Decision API and MCP will be built over one deterministic core in a separate release.