app.merces.ai

Risk Score

Know an agent before you deal with it

Paste an agent id, an owner address, a service URL or a name. The check reports what the registry knows about the agent as risk factors, and says unknown where it knows nothing.

Examples: base:25975, an owner address starting with 0x, capminal.ai, or a name.

What the check reports

Five groups of factors

  • Identity

    What the chain says about who this agent is.
  • Registration file

    The profile the agent publishes about itself.
  • Reputation

    On-chain reviews and validator checks.
  • Ownership

    Who controls the agent and whether that changed.
  • Registration pattern

    Whether the agent is one of many produced the same way.

The same check, one HTTP call

Call it before your agent pays a stranger, or before you let a stranger's agent check out.

API documentation

How a factor reads

  • OK. Present and in good shape: a readable file, reviews from many addresses, no ownership change.
  • Caution. Worth a second look: a very new identity, a thin profile, reviews from one address, a bulk-registered owner.
  • Warning. Something is broken or explicitly negative: an unreachable file, a declared inactive agent.
  • Info. A fact without a verdict: declared protocols, trust models, a payment wallet.
  • Unknown. We know nothing. We say so instead of guessing.

No number, yet

There is no scoring model behind the check today. Every factor is a fact from the registry: the chain record, the registration file, on-chain reviews and validator responses, and whether the agent is one of thousands produced the same way. A number arrives once it can be trained on verified outcomes, and it will be versioned so you can pin it.