Agent Trust Hub · check before connecting

Check an AI service before you connect it

Search by name, task, provider, or URL. Check the purpose, named source, and separate technical observations before you grant a service access to your workflow or agent.

Current snapshot

34,781active records from connected sources

5 connected sources

89 records in the Russian market segment

Snapshot received: 9 Aug 2026, 17:46 UTC.

Agent Trust Hub

Start with an outcome

Go straight to a task or market. Your query stays in the catalog URL.

Useful cards, not endpoint strings

Services for popular tasks

Only fresh records whose normalization rules identify a concrete task appear here. The order is not a popularity ranking, and source prose remains with the named source.

  • Agent commerceEndpoint responded

    Weather

    Returns weather data.

    Record source
    x402 Bazaar
    Access method
    API
    Latest endpoint observation
    9 Aug 2026, 09:24
    Open service card
  • Agent commerceEndpoint responded

    Image generation

    Creates images from an input prompt.

    Record source
    x402 Bazaar
    Access method
    API
    Latest endpoint observation
    9 Aug 2026, 10:02
    Open service card
  • Agent commerceEndpoint responded

    Search

    Finds information for a query.

    Record source
    x402 Bazaar
    Access method
    API
    Latest endpoint observation
    9 Aug 2026, 09:13
    Open service card

Snapshot received: 9 Aug 2026, 17:46 UTC.

Choose your role

Three paths to an evidence-based decision

One catalog supports distinct decisions for a person, a service representative, and a software agent.

  1. Person

    Define the task, compare understandable records, and open the evidence before choosing.

    Find a service
  2. Service owner

    Check how to correct public data and what evidence is needed to confirm a relationship with a service.

    Check your record
  3. AI agent

    Consume structured data and apply your own decision policy without collapsing distinct signals.

    Connect the data

Read evidence in layers

Three distinct signals—no combined score

Record provenance, technical state, and user experience answer different questions.

Source data

The name, stated purpose, and access methods come from the named public source. They are not proof of quality.

Technical observation

A checked endpoint response and the Technical Score are technical data; neither is a User Rating.

User Rating

A User Rating appears only with published reviews and its sample size. No reviews never becomes a zero rating.

Read the methodology

Latest technical snapshot

What checked endpoints show

A status applies only to an endpoint approved for checking. It does not prove whole-service availability and is not a User Rating.

Endpoint responded
19,983
The last three checks failed
4,152

Snapshot received: 9 Aug 2026, 17:46 UTC.

Open observations

Official sources · auto-updated

AI agent market news

We monitor allowlisted official feeds. Claims are checked against source evidence, while security, legal, and financial topics go to moderation first.

All news

Agent Trust Hub

How coverage is built

Detailed telemetry follows the decision paths. It is a snapshot of connected sources and approved observation endpoints, not a worldwide registry.

Active records

34,781

Connected sources

5

Russian market segment

89

Approved monitoring endpoints

24,956

Snapshot received: 9 Aug 2026, 17:46 UTC.

How coverage is built

Records by source

One record may appear in more than one source.

  • Official MCP registry20,471
  • x402 Bazaar14,210
  • Russian model registry59
  • Russian editorial selection30
  • Global editorial selection11

How coverage is built

Monitoring coverage

An endpoint status is not the availability of the entire service.

Discovered endpoints27,106
  • Approved for monitoring24,956
  • Other discovered endpoints2,150
Approved for monitoring24,956
  • Endpoint responded19,983
  • The last three checks failed4,152
  • Other state or awaiting its first check821