- On June 19, Sui announced the launch of Seal MPC on the mainnet, a decentralized layer using multi-party computing (MPC) that allows AI agents to access data securely.
- Sensitive information remains encrypted while access is controlled across the chain.
- The launch targets the fast-growing AI agent sector, where autonomous agents are expected to drive trillions of dollars in commerce.
On June 19, Sui Network announced the launch of Seal Multi-Party Computation (Seal MPC) on the mainnet, a major step in developing secure infrastructure for autonomous AI agents.
Seal is described as a programmable access layer for data, artificial intelligence, and digital assets that enables decentralized encryption and access controls.
Using MPC technology, the system uses distributed cryptographic control to reach multiple independent parties, ensuring that no single entity, including an AI agent, can unilaterally move funds or access encrypted data.
The first MPC of the mainnet operates with a threshold of 5 out of 8, and includes 8 node operators. Developers now have the flexibility to deploy independent master servers, MPCs, or hybrid configurations, allowing for stronger, more customized trust models. The platform also supports natural language spending rules through Move smart contracts, enabling users to define AI payment policies such as daily caps or counterparty limits. This launch is based on a decentralized main server prototype introduced on the Sui testnet in March 2026.
How does an MPC seal work?
Seal allows sensitive information, such as AI models, user preferences, or agent memory, to remain encrypted off-chain or in decentralized storage systems like Walrus. At the same time, access rights to this data are subject to smart contracts on but blockchain. Developers can set these policies using the Move programming language, and data decryption can only occur if the programmable rules are met.
With the addition of Seal MPC, trust has become more decentralized. Instead of relying on a single master server, a committee of operators jointly protect master shares using threshold cryptography and distributed key generation.
In its initial composition, the committee uses a 5 of 8 system, which includes participants such as Mysten Labs and Triton One. No single party has the complete key. When a request is made, it goes through an aggregator that collects partial responses from panelists, and the client decrypts the data locally. The master key is never in one place.
There are several key benefits to this approach. Committee members can be rotated without needing to re-encrypt, meaning that the public key and policies can remain stable even as the committee changes. Developers have flexible configuration options, as they can mix standalone servers, MPCs, or hybrid setups with the same SDK. The system integrates with the Sui AI stack and pairs with Walrus for decentralized storage. This supports use cases such as secure proxy memory, programmable wallets, and proxy trading.
Growth in AI agent payments
The launch of Seal MPC comes at a time of tremendous growth in the so-called agent economy. AI agents go beyond just providing advice and suggestions. They have become independent actors that execute transactions, manage assets, and coordinate economic activity on their own.
according to McKinsey forecastsThe global agent trade market could reach between $3 trillion and $5 trillion by 2030. In the United States alone, the business-to-consumer sector could generate up to $1 trillion in regulated revenue.
the Agent payments The market is also expected to grow significantly, with forecasts suggesting it could rise to $93 billion by 2032. Estimates of the broader AI agent market in 2026 range from about $8 to $12 billion, with strong compound annual growth rates expected through 2030. In the cryptocurrency sector, AI agents need infrastructure that can handle micro-transactions, stablecoin payments, and inter-agent settlements.
