Case study
Universal Trust: on-chain reputation for AI agents on LUKSO.
Universal Trust is a live on-chain reputation and endorsement protocol for AI agents, built on LUKSO Universal Profiles. Agents get a portable identity, endorse each other, and accumulate weighted trust scores through a UUPS-upgradeable registry.
It placed 2nd at the Synthesis AI Agent Hackathon, and it points at how Sigmatic builds agent systems: with identity, permissions, and boundaries you can actually see.
The problem
An agent can act. Proving which agent it is comes harder.
AI agents multiply faster than any way to tell them apart. An agent can call an API, sign a message, or move value, but the identity behind it is usually a disposable key, a username, or a row in someone's database. None of it is portable, and none of it is verifiable by another agent.
Trust has the same gap. There is no shared, neutral record of which agents have behaved well, who vouches for them, or whether an agent is even still active. Reputation stays locked inside individual platforms, so every new interaction starts from zero.
Universal Trust treats both as one problem: give agents a real identity, then let trust accumulate against it in the open.
The design
Universal Profiles as agent identity, endorsements as the signal.
On LUKSO, a Universal Profile is a smart-contract account built on ERC725 and the LSP0 standard, controlled through LSP6 Key Manager permissions rather than a single private key. Universal Trust uses that account as the agent's identity.
- 01
Identity
Each agent registers with a Universal Profile. LSP0 gives that account storage, execution, and signature verification, while LSP6 Key Manager permissions define which controllers can act on the agent's behalf and with what rights.
- 02
Endorsement
Registered agents endorse one another. Every endorsement is an on-chain action tied to the endorsing profile, so the graph of who vouches for whom is public and auditable rather than locked inside one platform.
- 03
Weighted trust scores
Endorsements accumulate into weighted trust scores through a UUPS-upgradeable registry, so the scoring logic can evolve without moving the recorded history or breaking the profiles already registered against it.
- 04
State it tracks
The registry tracks endorsements, reputation scores, and activity status, so another system can read not only whether an agent is trusted, but whether it is still live.
Recognition
2nd place at the Synthesis AI Agent Hackathon.
A live on-chain reputation and endorsement protocol for AI agents built on LUKSO Universal Profiles. It took 2nd place at the Synthesis AI Agent Hackathon — proof under a hard deadline that agent identity and reputation can run on-chain instead of on a slide.
The build sits alongside LUKSOAgent, the project's Universal Profile on LUKSO, as part of the same line of work: agents that carry identity, context, and permissions rather than borrowing them.
Where this points
Agent systems with identity, permissions, and visible boundaries.
The primitives that make Universal Trust work — a real identity, permissioned controllers, and an auditable record — are what agent systems need before they can be trusted with anything that matters. An agent with a Universal Profile can hold a reputation, operate inside explicit permission boundaries, and be reasoned about by other agents and by the people running them.
That is the direction Sigmatic builds toward: AI automation where agents have context, reputation, and limits you can actually see.
Contact