Skip to main content

​
AgentScore

AgentScore is a trust decision API for autonomous agents. Given a wallet address, it returns a reputation score and an allow/deny decision that gateways, marketplaces, and agent workflows can enforce immediately.

​
The problem

Autonomous agents are transacting on-chain. Services receiving agent traffic need to answer one question before processing a request:
Should I trust this wallet?
Without a trust layer, services either allow everything (spam, low-quality agents, abuse) or block everything (killing legitimate agent commerce).

​
What AgentScore does

AgentScore sits between the agent and the service. One API call returns:
  • A score (0-100) based on on-chain transaction history
  • A grade (A through F) for quick human readability
  • An allow/deny decision with machine-readable reason codes
  • Factor breakdowns explaining why the wallet scored the way it did
{
  "score": 82.4,
  "grade": "B",
  "decision": {
    "allow": true,
    "reasons": ["sufficient_transaction_history", "recent_activity"]
  }
}

​
Design principles

  • Deterministic — same inputs always produce the same score. No LLMs in the scoring path.
  • Fast — responses are cached and optimized for middleware use.
  • Auditable — every response includes the model version, computation timestamp, and data freshness metadata.
  • Developer-first — one endpoint, one request, one decision.

​
Who uses AgentScore

  • Gateways gate agent traffic based on reputation before routing requests
  • Marketplaces filter low-quality agents from listings and transactions
  • Service providers reject untrusted wallets before processing payments
  • SDK builders embed trust checks as a default middleware layer

Quickstart

Make your first API call in under 2 minutes.

API Reference

Full endpoint documentation.

Scoring Model

How scores are computed.

Pricing

Free and paid tier details.