The intersection of artificial intelligence and decentralized finance (DeFi) has long promised a future of autonomous on-chain commerce. However, a persistent architectural bottleneck has hindered this vision: the security of private keys. Giving an autonomous software agent direct access to a private key introduces catastrophic security vulnerabilities, while denying them transactional authority renders them ineffective.
To resolve this fundamental tension, Mysten Labs has officially launched Sui Seal MPC on the Sui mainnet. This cryptographic infrastructure is designed to enable autonomous AI agents to execute complex, policy-bound on-chain transactions without ever directly holding, exposing, or controlling full private keys. By combining multi-party computation (MPC), distributed key shares, and the native safety features of the Move programming language, Sui Seal MPC establishes a secure framework for the burgeoning "agentic web."
1. Main Facts: The Architecture of Sui Seal MPC
At its core, Sui Seal MPC is a security and transaction-signing framework designed to mitigate the risks associated with software-led asset management. In traditional Web3 setups, any entity—human or software—wishing to interact with a smart contract must sign transactions using a private key. If an AI agent operates using a traditional hot wallet, the entire private key must reside in the agent’s memory or hosting environment. If that environment is compromised via a server hack, dependency exploit, or LLM (Large Language Model) "prompt injection" attack, the key can be stolen, and the wallet’s treasury drained instantly.
传统 Hot Wallet 模式:
[AI Agent] ---> 拥有完整 [Private Key] ---> 直接签署交易 (高风险:一旦被黑,资金全无)
Sui Seal MPC 模式:
[AI Agent] ---> 触发交易请求
|
v
[Move Smart Contract Policies] (检查:每日限额、白名单、交易对手)
| (通过验证)
v
[Distributed MPC Nodes] (节点 1, 节点 2, ... 节点 N 协同计算分片)
|
v
[On-Chain Execution] (不暴露完整私钥,安全完成交易)
Sui Seal MPC solves this vulnerability through a three-pillared architectural approach:
- Multi-Party Computation (MPC): Private keys are mathematically split into "key shares" and distributed across multiple independent nodes. No single node ever possesses or knows the complete private key.
- Move-Based Policy Enforcement: Transactions are not merely signed; they are validated against strict, immutable rules written directly into Move smart contracts. These policies act as non-custodial guardrails, defining exactly what an agent can and cannot do.
- Cryptographic "Hidden Bids": The infrastructure supports advanced cryptographic primitives that allow AI agents to submit encrypted bids to decentralized markets. These bids remain hidden from public mempools until a synchronized, automated reveal phase, neutralizing front-running and sandwich attacks.
By decentralizing the signing authority and subjecting every transaction to programmatic on-chain compliance, Sui Seal MPC ensures that even if an AI agent’s frontend or decision-making logic is fully compromised, the underlying capital remains shielded behind strict cryptographic and smart-contract-defined boundaries.
2. Chronology: The Road to Agentic Web3
The launch of Sui Seal MPC is the culmination of several years of development at the intersection of cryptography, smart contract design, and artificial intelligence. Understanding how the industry arrived at this milestone requires tracing the evolution of automated on-chain execution.
+-----------------------------------------------------------------+
| 2015–2020: Early Bot Era |
| Simple, rigid scripts (arbitrage, liquidations). |
| High security risks; keys stored in plain text configuration. |
+-----------------------------------------------------------------+
|
v
+-----------------------------------------------------------------+
| 2021–2023: Rise of MPC & Multi-Sig |
| Institutional custody adopts MPC. |
| However, systems remain human-centric, requiring manual approval|
| or rigid API integrations unfit for dynamic AI agents. |
+-----------------------------------------------------------------+
|
v
+-----------------------------------------------------------------+
| 2024: The Agentic Web Boom |
| LLMs and AI agents enter DeFi. |
| The "Private Key Dilemma" becomes a critical bottleneck. |
+-----------------------------------------------------------------+
|
v
+-----------------------------------------------------------------+
| Q1 2025: Sui Seal MPC Launch |
| Mysten Labs deploys native MPC infrastructure on Sui mainnet. |
| Combines MPC signing with Move smart-contract policy engines. |
+-----------------------------------------------------------------+
The Early Bot Era (2015–2020)
In the early days of decentralized finance, automation was limited to simple, deterministic scripts (e.g., arbitrage bots, liquidation keepers). These scripts ran on centralized servers, keeping private keys in plain-text environment variables or local databases. Because these bots performed highly repetitive, low-variance tasks, their attack surfaces were relatively small, though security breaches remained common.
The Rise of MPC and Multi-Sig (2021–2023)
As institutional capital entered the Web3 space, the industry migrated toward Multi-Party Computation (MPC) and multi-signature wallets (like Gnosis Safe) for asset protection. However, these systems were designed primarily for human-centric workflows, requiring multiple human key-holders to manually sign off on transactions or relying on rigid API endpoints that could not easily adapt to the dynamic decision-making of AI agents.
The Agentic Web Boom (2024)
With the rapid advancement of Large Language Models (LLMs) and autonomous agent frameworks (such as LangChain and AutoGPT), developers began building agents capable of reading on-chain data, analyzing market sentiment, and formulating complex trading or purchasing strategies. However, the "Private Key Dilemma" became a hard bottleneck: developers had to choose between complete autonomy (giving the agent a hot wallet, risking total loss) or manual human-in-the-loop verification (destroying the agent’s speed and autonomy).
The Launch of Sui Seal MPC (Q1 2025)
Recognizing this architectural gap, Mysten Labs developed Sui Seal MPC. The system was designed to bridge the gap between autonomous AI decision-making and secure, trustless execution. Following successful testnet validation and cryptographic audits, the infrastructure was officially deployed to the Sui mainnet, offering a native, production-ready solution for developers building autonomous on-chain applications.
3. Technical Deep Dive: How Sui Seal MPC Works
To understand the security guarantees of Sui Seal MPC, it is necessary to examine its underlying cryptographic and programmatic components. The system operates as a hybrid layer, marrying off-chain cryptographic computation with on-chain state verification.
+-----------------------------------------------------------------------+
| AI AGENT WORKFLOW |
| |
| 1. AI Agent generates transaction payload (e.g., Swap 100 SUI to USDC)|
+-----------------------------------------------------------------------+
|
v
+-----------------------------------------------------------------------+
| MOVE SMART CONTRACT |
| |
| 2. Policy Check: |
| - Is 100 SUI under the daily limit (500 SUI)? -> YES |
| - Is the target pool whitelisted? -> YES |
+-----------------------------------------------------------------------+
|
v
+-----------------------------------------------------------------------+
| SUI SEAL MPC NODES |
| |
| 3. Key-share nodes perform distributed computation. |
| 4. Generate valid cryptographic signature without assembling full key|
+-----------------------------------------------------------------------+
|
v
+-----------------------------------------------------------------------+
| SUI MAINNET |
| |
| 5. Transaction executed on-chain. |
+-----------------------------------------------------------------------+
Distributed Key Generation (DKG) and Threshold Cryptography
Under a standard signature scheme (like ECDSA or Ed25519), a transaction requires a single signature generated by a single private key. Sui Seal MPC replaces this single-point-of-failure model with a Threshold Signature Scheme (TSS) powered by Multi-Party Computation.
During the setup phase, a Distributed Key Generation (DKG) protocol is executed among a set of independent, trust-minimized nodes. This protocol generates a public key on-chain, while the corresponding private key is mathematically fragmented into mathematical "shares."
$$S_1, S_2, dots, S_n$$
These shares are distributed across the participating nodes.
When an AI agent initiates a transaction, it requests a signature from the MPC network. The nodes collaborate to generate a valid cryptographic signature using their individual shares. This process occurs via secure, multi-party cryptographic protocols, ensuring that:
- No single node ever transmits its key share to another node.
- The full private key is never assembled in any single memory location, server, or database.
- A signature can only be generated if a pre-defined threshold of nodes (e.g., $t$ out of $n$) successfully compute their shares.
Programmatic Guardrails via Move Smart Contracts
While MPC secures the private key from physical or digital theft, it does not prevent an AI agent from making poor decisions, falling victim to prompt injection, or suffering from "hallucinations" that result in unintended transactions. To address this, Sui Seal MPC couples cryptographic signing with the native programming model of the Move language.

Move is an object-centric, resource-oriented programming language designed for secure asset management. On Sui, assets are treated as first-class "objects" with strictly defined ownership and permissions. Under Sui Seal MPC, developers can deploy Move smart contracts that act as deterministic policy engines. Every transaction request submitted to the MPC nodes must first pass through these on-chain policies.
These policies can enforce highly granular constraints, including:
| Policy Type | Description | Use Case Example |
|---|---|---|
| Spending Caps | Restricts the maximum volume of assets an agent can transfer within a specific timeframe. | An agent is limited to spending no more than 500 SUI per 24-hour window. |
| Whitelisted Counterparties | Limits transaction execution to a pre-approved list of smart contract addresses or protocols. | An agent can only interact with verified liquidity pools on Cetus or DeepBook, preventing interaction with malicious contracts. |
| Asset Restrictions | Restricts the types of tokens or assets the agent is authorized to trade, hold, or transfer. | A treasury rebalancing agent is allowed to trade only SUI, USDC, and USDT. |
| Temporal Locks | Restricts transaction execution to specific hours or enforces mandatory cooling-off periods. | Preventing high-frequency trading loops from draining liquidity during volatile market anomalies. |
If an AI agent is manipulated into requesting a transaction that violates any of these Move-defined rules, the smart contract immediately aborts the transaction, and the MPC nodes will refuse to sign the payload, neutralizing the threat.
Mitigating Front-Running via "Hidden Bids"
In decentralized markets, public mempools allow malicious actors, searchers, and MEV (Maximal Extractable Value) bots to view pending transactions and execute front-running or sandwich attacks. Because AI agents rely on automated, data-driven algorithms, their trading patterns can be highly predictable, making them prime targets for exploitation.
To counter this, Sui Seal MPC introduces support for "hidden bids." Using this mechanism, an AI agent can generate and submit an encrypted transaction payload representing a market bid or order. The transaction is recorded on-chain in its encrypted state, hiding the bid parameters (such as price, asset, and volume) from public view.
At a predetermined block height or epoch, a synchronized cryptographic reveal protocol is executed by the MPC nodes, decrypting and executing the bid simultaneously. This prevents front-running bots from reading the agent’s intent in the mempool and manipulating market prices ahead of execution.
4. Market Context and Industry Perspectives
The launch of Sui Seal MPC occurs amid a broader industry push toward "Agentic Web3"—a paradigm shift where the primary users of blockchain networks are expected to transition from human retail traders to autonomous software agents.
Several prominent Layer-1 and Layer-2 networks are currently vying for dominance in this emerging sector. For example, Solana has seen a surge in AI-themed memecoins and primitive agent integrations, while Near Protocol has championed "Chain Abstraction" to facilitate cross-chain agent operations. However, many of these existing implementations rely on centralized custodial APIs or off-chain MPC providers that introduce latency and trust assumptions.
Industry experts emphasize that the success of agentic applications will depend on native, low-latency infrastructure. In a recent technical discussion on blockchain scaling, cryptography researchers noted:
"The bottleneck for AI agents on-chain has never been the AI models themselves; it is the trust architecture. If an agent has to wait for a centralized custodian to approve a transaction, or if it has to risk its entire treasury on a hot-wallet key stored in an AWS instance, enterprise adoption remains impossible. Native, smart-contract-integrated MPC is the logical path forward."
By integrating MPC directly at the network and smart-contract level, Mysten Labs is attempting to establish Sui as the default execution layer for autonomous agents. This strategy leverages Sui’s high throughput (exceeding 100,000 TPS in testing), sub-second finality, and low, predictable gas fees—all of which are critical for high-frequency, agent-to-agent (A2A) commerce.
5. Strategic Implications and the Road Ahead
The deployment of Sui Seal MPC on the mainnet has significant implications for developers, enterprise institutions, and the broader decentralized economy.
+-----------------------------------------------------------------------------------+
| SUI SEAL MPC IMPLICATIONS |
+-----------------------------------------------------------------------------------+
|
+-------------------------------+-------------------------------+
| | |
v v v
+------------------+ +------------------+ +-------------------+
| FOR DEVELOPERS | | FOR ENTERPRISES | | FOR DEFI |
| | | | | |
| Low-friction SDKs| | Secure B2B flow | | MEV-resistant bids|
| Move integration | | Risk-free agents | | Dynamic liquidity |
+------------------+ +------------------+ +-------------------+
For Web3 Developers
For developers building on Sui, Seal MPC lowers the barrier to entry for creating secure, autonomous applications. Rather than building custom cryptographic signing services or relying on expensive third-party custody APIs, developers can leverage native SDKs to implement MPC-based agent wallets. The tight integration with Move means that writing complex, state-aware security policies requires minimal boilerplate code, allowing developers to focus on refining their agents’ AI logic and utility.
For Enterprise Adoption
Enterprise integration of blockchain technology has historically been slowed by compliance and custody concerns. Corporate treasuries cannot tolerate the risk of a software bug or external hack draining a wallet. Sui Seal MPC provides the exact framework enterprises require: a system where an AI agent can be tasked with automated supply chain payments, treasury rebalancing, or marketplace procurement, but where the maximum financial exposure is strictly capped by immutable on-chain rules.
For the Evolution of DeFi
As AI agents become more prevalent, the nature of decentralized liquidity is likely to change. Automated agents can continuously monitor yields, funding rates, and liquidity imbalances across multiple protocols, executing rebalancing strategies far faster than human operators. By providing a secure, MEV-resistant execution environment via MPC signing and hidden bids, Sui is positioning its DeFi ecosystem (including DeepBook and native lending protocols) to capture a significant portion of this automated, algorithmic volume.
Challenges and Future Outlook
Despite the promise of Sui Seal MPC, the path to widespread adoption is not without hurdles. The system’s success will depend on:
- Developer Tooling: Providing clear, accessible SDKs and templates for developers to easily write and deploy Move-based policy contracts.
- User Education: Ensuring that developers and users understand the limits of MPC. While it secures the private key, it does not magically prevent an agent from making bad trading decisions if its underlying LLM is poorly calibrated.
- Decentralization of MPC Nodes: Maintaining a robust, distributed, and trust-minimized set of node operators to perform the MPC calculations, ensuring the network remains resilient against censorship and single points of failure.
Sui Seal MPC represents a highly practical infrastructure milestone at the intersection of AI and cryptography. By decoupling transaction authorization from private key exposure and securing both behind Move-based smart contract guardrails, Mysten Labs has delivered a foundational tool for the next generation of autonomous, on-chain commerce.
