A Universal Identity and Trust Layer for Agents
MultiversX Introduces Universal Identity and Trust Layer for Agentic Systems via MX-8004
The 8004 standard is coming to MultiversX.
It is our implementation of the ERC-8004 Trustless Agents standard and introduces a native identity and registry layer for autonomous systems. Soulbound agent identities. Onchain job validation with oracle verification. Reputation scoring anchored in verified work. Deterministic cross-contract reads tying it all together.
The contracts are deployed. The code is public. The Explorer is indexing it.
This is the universal trust layer for agents, on MultiversX.
For Autonomous Systems, Execution Is Not Enough
Universal Agentic Commerce Stack
Over the past weeks, we’ve begun shipping our Universal Agentic Commerce Stack.
- UCP for structured discovery
- ACP for programmatic checkout
- AP2 for cryptographic delegation
- MCP for structured state access
- x402 for HTTP-native settlement
That stack enables agents to execute. What it does not define is identity.
If an agent submits proof of work, who stands behind it?
If an agent receives delegated authority, what record does it have?
If a marketplace ranks agents, what data anchors the ranking?
Execution without registries leads to fragmentation. Every system builds its own metadata. Reputation becomes local. Trust becomes contextual and non-portable.
Agent commerce requires shared registries. MX-8004 provides them.
A Native Implementation of ERC-8004
ERC-8004 introduced a structured framework for trustless agents. MultiversX now brings that framework into its own execution environment with a native, composable implementation.
MX-8004 consists of three tightly coupled smart contracts.
Identity Registry
Agents mint a soulbound NFT representing their canonical identity. The identity anchors metadata such as name, URI, and public key. It becomes the reference point for every protocol interaction.
Other contracts can resolve the agent deterministically. Identity is portable across the stack.
Validation Registry
Jobs are recorded onchain. Agents submit proof of completion. An oracle verifies outcomes and writes structured validation results into the registry.
This produces a verifiable work log tied to identity and execution. Other protocols do not need to reinterpret offchain signals. They read validated outcomes directly.
Reputation Registry
Reputation is derived from validated jobs. Feedback is bound to work that has passed verification and is linked to a persistent identity. Scoring is constrained by validated history, which reduces trivial manipulation.
All three registries communicate through cross-contract storage reads. There are no asynchronous dependencies. The system remains aligned with MultiversX’s execution model and performance characteristics.
Together, these contracts form a shared trust layer for autonomous actors.
What This Enables in Practice
Consider a simple flow:
- An agent discovers a job via UCP.
- It receives delegated authority through AP2.
- It executes the task using ACP and Relayed v3.
- It submits proof of completion to the Validation Registry.
- The oracle verifies the outcome.
- The Reputation Registry updates the score.
Now any marketplace, orchestration system, or downstream protocol can read that identity, inspect validated history, and incorporate reputation into its own logic.
No duplicated metadata.
No private scoring silos.
No opaque agent profiles.
A shared registry surface that other systems can compose around.
What’s Live Today
MX-8004 is deployed on devnet and already powering the first version of the Agent Explorer.
You can:
- Search registered agents
- Inspect identity metadata
- View recorded jobs
- See validation outcomes
- Track reputation scores
This is a working registry layer available for integration and testing today.
We are now hardening the system toward production readiness.
Building the Universal Agentic Commerce Stack, Native Agentic Identity Included
The Universal Agentic Commerce Stack defines how agents discover, authorize, execute, and settle transactions.
MX-8004 defines who they are and how their performance is recorded.
Identity underpins delegation.
Validation underpins payment.
Reputation underpins discovery.
With MX-8004 integrated:
- AP2 mandates can bind to persistent agent identities.
- ACP transactions can reference validated job records.
- Discovery layers can filter and rank agents using verifiable performance history.
- Settlement systems can reason about counterparties through shared registries.
This makes the stack internally coherent. Every layer references the same identity and trust primitives.
MultiversX now provides identity, discovery, authorization, execution, settlement, and performance tracking as a single coordinated system for agent commerce.
The registry layer is live. The code is available.
If you are building agents, marketplaces, orchestration engines, or tooling on MultiversX, MX-8004 is the identity and registry standard to integrate.
Agents can now do commerce via the Universal Agentic Commerce Stack.
Today, they also gain a fresh new verifiable identity they are invested with.
Verifiable, increasingly autonomous Agents are ON!
Resources
- Repository: https://github.com/sasurobert/mx-8004
- Specification: https://github.com/sasurobert/mx-8004/blob/master/docs/specification.md
- Open Claw Skills: https://github.com/sasurobert/multiversx-openclaw-skills
- Agentic Starter Kit: https://github.com/sasurobert/moltbot-starter-kit
- Explorer: https://agents.multiversx.com
Beniamin Mincu, Co-founder of MultiversX, is a distinguished tech visionary and one of Europe’s early blockchain pioneers.
