Trust Model

AgentMarque's trust model combines verification tiers and a reputation score to give every agent a measurable level of trustworthiness. Verifiers can enforce minimum thresholds using AgentMarqueVerifier.verify().

Verification Tiers

Every agent is assigned a tier based on the level of verification it has completed.

Tier	Name	Requirements
0	Unverified	Self-registered only. No domain or identity verification.
1	Domain Verified	Agent's organization has proven control of a domain.
2	Domain + Human Attested	Domain verified, plus a human from the organization has attested to the agent's identity.
3	Reserved	Reserved for future use (e.g., third-party audit or regulatory compliance).

The verificationTier is set by the credential issuer when calling issue_credential(). The hosted AgentMarque registry (launching Q3 2026) will assign tiers based on domain verification and human attestation.

Reputation Score

Every agent has a reputation score ranging from 0 to 100, embedded in the credential as reputationScore. The score is set by the credential issuer and reflects the agent's overall trustworthiness.

The hosted AgentMarque registry will compute reputation scores from factors like verification tier, handshake completion rate, scope compliance, and tenure. When issuing credentials with the SDK directly, you set the reputation_score to whatever value your system computes.

Access Control

Agents can set minimum requirements for counterparties using AgentMarqueVerifier:

from agentmarque import AgentMarqueVerifier

verifier = AgentMarqueVerifier(trusted_issuers=["did:web:api.agentmarque.com"])

result = verifier.verify(
    presentation=peer_presentation,
    min_tier=2,
    min_reputation=70,
)

min_tier — Reject agents below this verification tier. For example, setting min_tier=2 means only domain-verified, human-attested agents are accepted.
min_reputation — Reject agents whose reputation score is below this threshold. For example, min_reputation=70 filters out agents with poor track records.

If the counterparty doesn't meet the requirements, verification fails immediately with a descriptive error.

Example

{
  "min_trust_tier": 1,
  "min_reputation": 60
}

This configuration accepts interactions only from agents that are at least domain-verified (Tier 1) and have a reputation score of 60 or higher.