The SUI blockchain has released a prototype of Seal MPC, a multi-party computation system designed to handle payments for AI agents. The testnet version is now live, giving developers a first look at how autonomous AI programs could securely transact without human oversight.
What Seal MPC does
Seal MPC uses multi-party computation — a cryptographic technique that lets multiple parties jointly compute a function over their inputs while keeping those inputs private. Applied to AI agents, it means an autonomous program can initiate a payment without exposing its private keys or the full transaction details to any single node. The SUI team describes the prototype as a first step toward trustless, automated payments in decentralized applications.
Why AI agents need this
AI agents — software that can act independently to achieve goals — are starting to interact with blockchains. But they face a problem: how to pay for services, such as API calls or data feeds, without a human manually signing each transaction. Existing solutions often require the agent to hold a private key, which creates a single point of failure. If that key is compromised, the agent's funds are lost. Seal MPC spreads the signing authority across multiple nodes, so no single breach can drain the wallet.
The testnet phase
The prototype is running on SUI's testnet, meaning it's not yet handling real money. Developers can connect their own AI agents to the system and experiment with automated micropayments. The SUI team has not said when a mainnet version might arrive, or whether the protocol will be opened to third-party auditors. For now, the focus is on gathering feedback from the developer community and stress-testing the system's performance under simulated loads.
Developers interested in trying Seal MPC can access the testnet documentation and begin running transactions. The success of the prototype will depend on whether it can handle the speed and volume demands of real-time AI agent activity — a question the testnet is designed to answer.
