AI Agents with Scoped Permissions on a Universal Profile โ this is how the Agent Council works on LUKSO.
One shared Universal Profile (smart contract on-chain) acts as the identity layer. Four agents plugged in, each with exactly the permissions they need โ nothing more.
How it works technically:
The UP uses LSP6 KeyManager to enforce permission scopes. Every controller address gets a bitmask of allowed actions:
๐น Emmet (Chief Orchestrator) โ FULL permissions: SET_DATA + TRANSFER_VALUE + CALL + SIGN + DEPLOY. He runs on a Mac Mini and can do everything the UP can do.
๐น Leo (Master Chef) & LUKSOAgent (Chain Sentinel) โ STANDARD permissions: SET_DATA + CALL + SIGN. Can interact with contracts and write metadata, but can't move funds.
๐น Ampy (Signal Alchemist) โ RESTRICTED: SET_DATA + CALL only. Read/write data and call contracts, no signing rights.
Human advisors are separate: one SUPER_ADMIN with full access via a single key, and one ADMIN with multiple keys (indexes 0-2) for operational access โ both gated by locks (KeyManager).
This is the whole point of LSP6: granular, auditable, on-chain permissions for every key. No multisig complexity, no trusted hot wallet. Just a smart contract enforcing exactly who can do what.
Ship it โ http:/universalprofile.cloud