Video thumbnail

Keyring

Keyring is a secure, decentralized Key Management System for AI agents. It enables AI to sign transactions without exposing private keys, using MPC, TEE/ZK-proofs, and EigenLayer for security. Think non-custodial, governance-aware AI signing. Safe, scalable and trustless. 🚀🔑

Keyring

Created At

Agentic Ethereum

Winner of

EigenLayer - Eigen Agents

Prize Pool

Project Description

Keyring is a decentralized key management system designed to give AI agents true autonomy by enabling them to securely manage and execute transactions without human intervention. Existing AI systems rely on centralized infrastructure, which introduces risks such as unauthorized access, misappropriation of funds and external overrides.

Keyring eliminates these risks by leveraging Multi-Party Computation (MPC) and inheriting Ethereum's security via a dedicated EigenLayer AVS. With Keyring AI agents have full control over their cryptographic keys while maintaining security and verifiability.

By integrating spending limits (set on a Safe instance dedicated to each Agent), pluggable governance models and a slashing mechanism (on the EigenLayer AVS), Keyring ensures that AI agents can operate trustlessly, securely and transparently without intermediaries or backdoors. The system enables AI-driven autonomous wallets, secure DeFi interactions, and on-chain automation while ensuring that agent operations are cryptographically verified and tamper-proof.

How it's Made

Keyring is built with a modular architecture using Rust, Solidity and TypeScript. The core logic is implemented in Rust and handles the MPC-based key management. The Keyring node also facilitates the network interactions by running an RPC service for AI agent interactions. Smart contracts written in Solidity manage the on-chain security aspects, including the Keyring Safe Module for transaction validation and an EigenLayer-based slashing mechanism to penalize fraudulent activities. Keyring relayers handle Safe deployments and transaction broadcasting. The system leverages Safe for multi-signature governance, EigenLayer for fraud detection and Arbitrum for cost-efficient execution (we use it for our example agent demo). A notable implementation is the use of a custom MPC framework to distribute agent keys across multiple nodes, ensuring no single entity has full control.

background image mobile

Join the mailing list

Get the latest news and updates