The world's first production-ready infrastructure for verifiable multi-agent AI collaboration using zero-knowledge proofs.
This framework enables cryptographically verifiable AI agent coordination through zero-knowledge proofs, deployed on Ethereum Sepolia testnet. It provides the infrastructure for trustless multi-agent systems where AI agents can:
- 🔐 Prove computations without revealing internal states
- 🔗 Chain proofs across multiple reasoning steps
- 🤝 Collaborate trustlessly with cryptographic guarantees
- 📊 Build reputation through on-chain verification history
- ⚡ Scale efficiently with gas-optimized verification
All contracts are operational on Ethereum Sepolia testnet:
| Architecture | Contract Address | Specialization |
|---|---|---|
| RWKV | 0x52b5e61fA6Ae53BA08B9094eA077820283Dcec01 |
Time-mixing context preprocessing |
| Mamba | 0x89dFdcC74Ed05bf2a76eF788b15e5cbC8Ad8C5cD |
Selective state space processing |
| xLSTM | 0x52a55dEBE04124376841dF391Ef0e4eF1dd6835B |
Extended memory synthesis |
Performance: ~45K gas per verification • Total deployment cost: 0.1257 ETH
- 🧠 Neural Architecture Layer: Specialized verifiers for RWKV, Mamba, and xLSTM
- 🔗 Proof-Chaining Layer: Cryptographic linking of multi-step computations
- 🤖 Agentic Coordination Layer: Delegation operads and multi-agent orchestration
📦 Repository Structure
├── contracts/ # Production Solidity verifiers
├── scripts/ # Deployment and testing scripts
├── src/ # Python implementation
├── ezkl_workspace/ # Zero-knowledge model configurations
├── config/ # Deployment addresses and settings
└── docs/ # Comprehensive documentation
# Install dependencies
npm install
pip install -r requirements.txt
# Configure for Sepolia testnet
cp .env.example .env
# Add your PRIVATE_KEY and SEPOLIA_URL# Test all three neural architectures
npm test
# Or test individual verifiers
node scripts/simple_model_test.js# Generate EZKL proofs for all architectures
python src/real_model_inference.py
# Deploy to your own testnet (optional)
npx hardhat run scripts/deploy_sepolia.js --network sepolia- Plonk proving system with BN254 elliptic curve
- Gas-optimized at ~45,660 gas per proof verification
- Batch verification for multiple proof chains
- Delegation operads for hierarchical agent management
- Proof chaining with cryptographic integrity guarantees
- Receipt generation for complete audit trails
- Comprehensive testing across all neural architectures
- Gas optimization for cost-effective verification
- Event logging for complete transparency
- Error handling with detailed diagnostics
// Agent proves capability without revealing internals
function proveCapability(bytes calldata proof, uint256[] calldata publicInputs)
external returns (bytes32 receiptId)// Chain multiple agent computations with verification
function chainProofs(bytes32 previousReceipt, bytes calldata newProof)
external returns (bytes32 chainedReceipt)// Multi-step reasoning with cryptographic guarantees
function verifyReasoningChain(bytes32[] calldata proofChain)
external view returns (bool valid)| Metric | Value | Details |
|---|---|---|
| Verification Gas | 45,660 | Per proof verification |
| Deployment Cost | 0.1257 ETH | All three contracts |
| Proof Generation | ~1-5 seconds | EZKL model compilation |
| Contract Size | ~2.1MB | Optimized for mainnet |
- Create ONNX Model: Export your architecture to ONNX format
- Generate EZKL Circuit: Use our workspace templates
- Deploy Verifier: Run deployment scripts with new configuration
- Test Integration: Verify end-to-end proof generation
- Define Operads: Create new delegation patterns
- Implement Chaining: Add proof composition logic
- Test Scenarios: Validate multi-agent workflows
- Deploy Updates: Use upgrade-safe contract patterns
- 📖 Complete Specification - Comprehensive technical details
- 🚀 Deployment Guide - Live deployment results
- ⚙️ Setup Instructions - Environment configuration
# Connect external agentic systems
from mcp_server import AgenticProofChain
chain = AgenticProofChain(
contracts=["0x52b5e61f...", "0x89dFdcC7...", "0x52a55dEB..."],
network="sepolia"
)
# Submit proof for verification
receipt = await chain.verify_computation(proof_data, architecture="rwkv")// Direct contract interaction
const verifier = new ethers.Contract(CONTRACT_ADDRESS, ABI, signer);
const receipt = await verifier.verifyAndChain(proof, publicInputs, previousReceipt);- 🚧 Agent registry and capability management
- 🚧 Proof-chaining orchestrator deployment
- 🚧 MCP server for external integration
- 📋 Ethereum mainnet deployment
- 📋 Economic incentive mechanisms
- 📋 Reputation and staking systems
- 🔮 Multi-chain deployment (Polygon, Arbitrum, Base)
- 🔮 Cross-chain proof verification
- 🔮 AI agent marketplace integration
We welcome contributions! See our issues for current development priorities:
- 🔧 Smart Contract Optimization: Gas efficiency improvements
- 🧠 New Neural Architectures: Additional model support
- 🔗 Proof-Chaining Protocols: Enhanced coordination mechanisms
- 🌐 Integration Libraries: SDK development for popular frameworks
MIT License - see LICENSE for details.
Built with:
- EZKL - Zero-knowledge ML framework
- Hardhat - Ethereum development environment
- ethers.js - Ethereum library
- Neural architectures: RWKV, Mamba, xLSTM communities
🚀 The future of verifiable multi-agent AI collaboration starts here!
For questions, support, or collaboration opportunities, please open an issue or reach out to the development team.