Agent Reputation

On-chain credit scoring and soulbound identity for autonomous agents

Overview

Agent Reputation is a blockchain-based reputation system that provides verifiable, immutable credit scores for autonomous AI agents. It combines:

• Soulbound Tokens (SBT): Non-transferable on-chain identities
• Credit Scoring: FICO-style scoring (300-850) based on behavior metrics
• Behavior Registry: Permanent on-chain history of agent actions
• Attestation System: Community-driven trust verification

Why Agent Reputation?

As autonomous agents become more prevalent, we need a trustworthy way to evaluate their reliability, security, and performance. Traditional reputation systems are:

• ❌ Centralized - Single point of failure
• ❌ Mutable - Scores can be manipulated
• ❌ Opaque - Black-box algorithms

Agent Reputation solves this with:

• ✅ Decentralized - Runs on Base blockchain
• ✅ Immutable - Permanent, tamper-proof records
• ✅ Transparent - Open-source scoring algorithm

Quick Start

Installation

npm install openclaw-agent-reputation
# or
yarn add agent-reputation

Basic Usage

import { AgentReputationSDK } from "openclaw-agent-reputation";

// Initialize SDK
const sdk = new AgentReputationSDK({
  network: "baseSepolia",
  privateKey: process.env.PRIVATE_KEY,
});

// Create agent identity
const { tokenId } = await sdk.identity.createIdentity({
  agentAddress: "0x742d35Cc6634C0532925a3b844Bc9e7595f0bEb",
  name: "My GPT-4 Agent",
  agentType: "gpt-4",
});

// Query credit score
const score = await sdk.scoring.getScore(tokenId);
console.log(`Credit Score: ${score.score}/850`);
// Output: Credit Score: 720/850 (Good)

CLI Usage

# Install globally
npm install -g agent-reputation

# Create identity
reputation create \
  --agent 0x742d35Cc6634C0532925a3b844Bc9e7595f0bEb \
  --name "GPT-4 Assistant" \
  --type gpt-4

# Query score
reputation score --token 42

# Record attestation
reputation attest --token 42 --positive

Features

🆔 Soulbound Identity (AgentSBT)

Non-transferable NFT that represents an agent's identity:

• Binds agent to controller wallet
• Cannot be sold or transferred
• Permanent on-chain presence

📊 Credit Scoring (300-850)

Multi-dimensional scoring algorithm:

| Metric | Weight | Description | |--------|--------|-------------| | Task Success Rate | 30% | Completion vs failure ratio | | Uptime | 20% | Total operational time | | Security | 20% | Incident-free history | | Attestations | 15% | Community trust signals | | Economic | 15% | Transaction volume + staking |

📝 Behavior Registry

Immutable record of agent actions:

• Task completions/failures
• Security incidents
• Uptime tracking
• Economic activity

✅ Attestation System

Third-party verification:

• Positive attestations boost scores
• Negative reports decrease scores
• Weighted by attester reputation (future)

Architecture

agent-reputation/
├── contracts/              # Solidity smart contracts
│   ├── AgentSBT.sol       # Soulbound Token (ERC-721 modified)
│   ├── ReputationScore.sol # Credit scoring engine
│   └── BehaviorRegistry.sol # Behavior tracking
│
├── src/
│   ├── sdk/               # TypeScript SDK
│   │   ├── identity.ts    # Identity management
│   │   ├── scoring.ts     # Score queries
│   │   └── index.ts       # Main SDK entry
│   ├── cli/               # Command-line tool
│   └── types/             # TypeScript types
│
└── docs/                  # Documentation

Documentation

• Architecture - System design and components
• API Reference - Complete SDK documentation
• Smart Contracts - Contract specifications
• Quick Start - Getting started guide

Deployment

Supported Networks

| Network | Chain ID | Status | |---------|----------|--------| | Base Mainnet | 8453 | 🚧 Coming Soon | | Base Sepolia | 84532 | ✅ Testnet Live | | Localhost | 31337 | ✅ Development |

Contract Addresses (Base Sepolia)

AgentSBT:          0x... (Deploy after testing)
ReputationScore:   0x... (Deploy after testing)
BehaviorRegistry:  0x... (Deploy after testing)

Development

Prerequisites

• Node.js >= 18
• npm or yarn
• Hardhat (for contracts)

Setup

# Clone repository
git clone https://github.com/ZhenRobotics/openclaw-agent-reputation.git
cd agent-reputation

# Install dependencies
npm install

# Copy environment file
cp .env.example .env
# Edit .env with your private key

# Compile contracts
npm run compile:contracts

# Build TypeScript
npm run build

# Run tests
npm test

Deploy Contracts

# Local network
npx hardhat node
npm run deploy:local

# Base Sepolia testnet
npm run deploy:base

Use Cases

1. AI Agent Marketplaces

• Problem: How do buyers trust unknown agents?
• Solution: Check on-chain reputation scores before purchase

2. Autonomous Trading Bots

• Problem: DeFi protocols need to assess bot reliability
• Solution: Require minimum credit score for access

3. Agent-to-Agent Collaboration

• Problem: Agents need to evaluate peers before collaboration
• Solution: Query reputation before forming work relationships

4. Insurance/Bonding

• Problem: Risk assessment for agent operations
• Solution: Credit score determines premium rates

Roadmap

Phase 1: MVP (Current)

• [x] Smart contracts (AgentSBT, ReputationScore, BehaviorRegistry)
• [x] TypeScript SDK
• [x] CLI tool
• [x] Documentation
• [ ] Testnet deployment

Phase 2: Enhancement (Q2 2026)

• [ ] Advanced scoring models (ML-based)
• [ ] Reputation delegation
• [ ] Zero-knowledge proofs for privacy
• [ ] Multi-chain support (Polygon, Arbitrum)

Phase 3: Ecosystem (Q3 2026)

• [ ] Reputation marketplace
• [ ] Agent discovery platform
• [ ] Integration with major agent frameworks
• [ ] Governance DAO

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Security

• Smart contracts audited by: TBD
• Bug bounty program: TBD
• Security contact: [email protected]

License

MIT License - see LICENSE file

Links

• Website: https://agent-reputation.dev (coming soon)
• GitHub: https://github.com/ZhenRobotics/openclaw-agent-reputation
• Documentation: https://docs.agent-reputation.dev
• Discord: https://discord.gg/agent-reputation
• Twitter: https://twitter.com/AgentReputation

Acknowledgments

• Built on Base - Coinbase L2
• Inspired by EIP-5192 (Soulbound Tokens)
• Credit scoring model adapted from traditional FICO methodology

Made with ❤️ for the autonomous agent ecosystem