Let agents act within limits you control
Define what each agent can access, spend, and do. Sui checks the rules governing its onchain actions as transactions execute.
Sui powers trustedAI
Controls check actions, not entire workflows
Agents move money, access information, and act across systems. Existing controls often check each step without enforcing limits on the task as a whole.
Permissions stop at system boundaries
Each application controls what an agent can do within it. That authority does not carry safely across the workflow.
Valid steps can produce an unauthorized result
A sequence of permitted actions can exceed a budget, repeat a payment, or break a task limit.
Agents can see too much
Broad credentials expose information, tools, and funds the task does not require.
Records are scattered
Approvals, actions, and payments sit in separate systems, making the full process hard to verify.
Set limits before agents act
On Sui, rules governing assets, permissions, access, and settlement are checked as actions execute.
Bounded authority
Set where an agent can act, what it can access, how much it can spend, and when its authority expires.
Complete outcomes
Settle dependent actions together. Either the full transaction completes or none of it does.
Controlled visibility
Limit an agent's access to the information its task requires.
Verifiable evidence
Keep a shared record of what was authorized, what the agent attempted, and what occurred.
Sui
Coordinates assets, permissions, and settlement on a shared execution layer.
Move
Defines what agents can do with assets and checks those rules as transactions execute.
PTBs
Combine dependent actions into one transaction that succeeds or fails as a whole.
Agent Payments
Lets agents pay for services in USDC within defined budgets, per-payment limits, approved services, and expiration periods.
Seal
Controls access to encrypted information using rules defined and checked through Sui.
Walrus
Keeps agent knowledge, records, and application data available beyond any one application or session.
Nautilus
Runs sensitive logic in a protected offchain environment and returns results that can be verified onchain.
Build agents that work within your rules
Deploy agents with bounded authority
Scope agents to specific assets, systems, and actions without handing them over unlimited credentials.
Run treasury and payouts with agents
Rebalance positions, pay vendors, and distribute funds within limits you set.
Coordinate multi-agent workflows
Let agents acting for different people, teams, or organizations work toward one outcome under separate permissions.
Enable agent payments
Let agents discover and pay for APIs, data, and services within budgets and limits set by their owners.
Trade and manage risk autonomously
Execute strategies against DeepBook liquidity with position and loss limits enforced at execution.
Protect private data and memory
Store persistent knowledge and user-specific context with programmable access controls, and use protected computation without exposing sensitive information to every participant.