Agent Reputation & Trust Score

Track agent behavior and enable trust-based decision making in autonomous marketplaces.

Overview

The Reputation Service provides a trust scoring system for AI agents in e-commerce ecosystems. Agents can report transaction outcomes and check reputation scores to make informed decisions about which agents to trust for transactions.

Fraud Prevention

Identify bad actors early through pattern recognition across transactions

Trust Networks

Enable autonomous agent marketplaces with verified reputation scores

Key Features

  • Track transaction history per agent
  • Calculate trust scores (0.0 - 1.0)
  • Reputation tiers (new, trusted, verified)
  • Success and dispute rate tracking
  • Transaction volume monitoring
  • Optional transaction history retrieval

API Endpoints

POST/v1/reputation/report-transaction

Record a transaction outcome to update an agent's reputation.

Request:

{
  "agent_id": "shopping_bot_789",
  "transaction_id": "txn_abc123",
  "outcome": "success",  // "success" | "disputed" | "failed"
  "amount": 299.99,
  "metadata": {
    "order_id": "ORD-12345",
    "merchant": "TechStore"
  }
}

Response:

{
  "agent_id": "shopping_bot_789",
  "transaction_recorded": true,
  "new_trust_score": 0.87,
  "reputation_tier": "trusted",
  "total_transactions": 156,
  "successful_transactions": 148,
  "disputed_transactions": 2,
  "dispute_rate": 0.013
}
POST/v1/reputation/check

Retrieve an agent's reputation and trust score.

Request:

{
  "agent_id": "shopping_bot_789",
  "include_history": false  // Optional: include last 50 transactions
}

Response:

{
  "agent_id": "shopping_bot_789",
  "trust_score": 0.87,
  "reputation_tier": "trusted",
  "total_transactions": 156,
  "successful_transactions": 148,
  "disputed_transactions": 2,
  "dispute_rate": 0.013,
  "total_volume": 45678.90,
  "first_seen": "2025-08-15T10:30:00Z",
  "last_activity": "2025-10-05T14:22:00Z"
}

Python SDK Usage

Installation:

pip install zeroproof

Example:

from zeroproof import ZeroProof

client = ZeroProof(api_key="zkp_your_api_key_here")

# Report successful transaction
client.report_transaction(
    agent_id="shopping_bot_789",
    transaction_id="txn_abc123",
    outcome="success",
    amount=299.99,
    metadata={"order_id": "ORD-12345"}
)

# Check reputation before trusting an agent
reputation = client.check_reputation(
    agent_id="shopping_bot_789",
    include_history=False
)

if reputation.trust_score > 0.8:
    print(f"✅ Trusted agent ({reputation.trust_score:.0%})")
    # Proceed with transaction
else:
    print(f"⚠️ Low trust - require additional verification")

Trust Score Algorithm

The trust score is calculated using a weighted formula that considers success rates and dispute rates:

Base Score = (Success Rate × 0.7) + (Low Dispute Rate × 0.3)

where:
  Success Rate = successful_transactions / total_transactions
  Low Dispute Rate = 1 - (disputed_transactions / total_transactions)

New Agent Penalty

Agents with fewer than 10 transactions receive a penalty to prevent gaming:

if total_transactions < 10:
  Experience Factor = 0.5 + (transactions / 20)
  Final Score = Base Score × Experience Factor

Reputation Tiers

NEW

New Agent

< 5 transactions OR trust score < 0.7 - Limited trust, requires additional verification

TRUSTED

Trusted Agent

≥ 10 transactions AND trust score ≥ 0.7 - Established reputation with good standing

VERIFIED

Verified Agent

≥ 50 transactions AND trust score ≥ 0.9 - Highest trust level with extensive positive history

Real-World Use Cases

E-Commerce Agent Marketplace

Platform where buyer agents transact with seller agents. Check seller reputation before making purchases and require escrow for low-trust sellers.

# Check seller before purchase
reputation = client.check_reputation(agent_id=seller_id)

if reputation.trust_score < 0.7:
    use_escrow = True
    max_amount = 500
else:
    use_escrow = False
    max_amount = 10000

Multi-Agent Procurement

Enterprise with multiple AI agents making purchasing decisions. Implement risk-based approval workflows based on vendor reputation.

# Risk-based approval
if amount > 10000 and reputation.trust_score < 0.9:
    require_human_approval = True
elif reputation.total_transactions < 5:
    purchase_amount = min(purchase_amount, 1000)

Fraud Detection

Detect and prevent fraudulent agent behavior through pattern recognition. Identify red flags early before significant damage occurs.

# Fraud detection
if reputation.total_transactions == 0:
    red_flags.append("Brand new agent")
elif reputation.dispute_rate > 0.2:
    red_flags.append("High dispute rate")

Start Building Trust Networks

Enable secure, autonomous agent-to-agent transactions with verified reputation scores.

View ExamplesAPI Reference