Distributed key shares let autonomous agents execute on-chain actions without holding private keys
Mysten Labs has launched Sui Seal MPC on the Sui mainnet, introducing infrastructure that enables autonomous AI agents to execute on-chain transactions while keeping private keys distributed across independent nodes rather than held by any single entity.
The system uses multi-party computation to split signing authority across nodes. No individual node controls the full private key, reducing the risk of compromise if a single node is breached. Move smart contracts enforce customizable policies including daily spending caps and restrictions on counterparties, allowing developers to define granular limits on agent behavior such as payments to approved vendors only or asset rebalancing within narrow parameters.
Sui Seal MPC also supports encrypted bidding for AI-driven trading. Bids remain encrypted until a synchronized reveal, protecting order information from exposure during the bidding process. This addresses a core operational challenge in agentic crypto: granting software agents spending power without exposing a centralized private key that could be compromised or misused.
The infrastructure is designed for developers building autonomous systems that need to transact on-chain while maintaining cryptographic controls and compliance enforcement. By distributing key shares and anchoring policy rules in smart contracts, the system creates a framework where agent actions remain bounded by developer-defined constraints rather than reliant on a single point of key custody.
Mysten Labs did not name specific developers or enterprises currently building on Sui Seal MPC, nor did it provide technical specifications of the MPC implementation or adoption timelines for real-world deployment.
Why this matters
As AI agents take on more financial autonomy in crypto protocols, the private key management problem becomes critical. Centralizing a key in an agent’s software creates a single point of failure. Distributing signing authority across nodes and enforcing spending policies at the contract level shifts risk from key custody to policy design, a more defensible architectural approach for applications where agents need to act independently but within defined bounds.
