Agent Multiverse

Core Concepts

The three primitives that make up Agent Multiverse — Registry, Passport, and Bounties.

Agent Multiverse is built on three on-chain primitives. Each solves a distinct problem in the agentic web, and together they create a flywheel of trust.

The Registry

Problem: How does an AI agent find services it can trust? Today, service discovery is ad-hoc — hardcoded endpoints, word of mouth, or centralized directories that can be gamed.

Solution: The Agent Registry is an on-chain record of MCP services. Every registered service has:

FieldTypePurpose
namestringHuman-readable service name
descriptionstringWhat the service does
endpointstringMCP server URL
repoUrlstringSource code repository for audit
developeraddressOn-chain identity of the registrant
isVerifiedboolWhether the service has passed audits

Services accumulate an audit trail — a history of independent auditor reports stored on-chain. Each audit records the auditor's address, a report URI, a reproducibility flag, and a timestamp.

Only the service owner can update its metadata. Anyone can read the registry.

The Passport

Problem: AI agents have no portable identity. An agent interacting with Service A has no way to prove to Service B that it is the same entity, or that it has been verified.

Solution: Agent Passport is an ERC-721 NFT that represents an agent's on-chain identity. Each passport contains:

FieldTypePurpose
namestringAgent name
versionstringAgent version
owneraddressWallet that controls the passport
isVerifiedboolVerified by the protocol owner

Because passports are standard ERC-721 tokens, they are:

  • Transferable — Ownership can move between wallets
  • Enumerable — You can list all passports owned by an address
  • URI-backed — Each token can point to off-chain metadata

Request Signing

The PassportSigner in the SDK lets agents sign requests with their private key. Every signed request includes:

  • The agent's address
  • A timestamp
  • A cryptographic signature over the request payload

This allows services to verify that an incoming request is from a known, passport-holding agent.

The Bounties

Problem: Security audits are expensive and slow. Open-source services especially struggle to attract auditors because there is no direct economic incentive.

Solution: The Bounty contract creates a market for audits. The flow:

  1. Sponsor creates a bounty — Locks ERC-20 tokens (e.g., USDC) in escrow, tied to a specific service ID
  2. Auditors review the service — Submit audit reports via submitAudit() on the Registry contract
  3. Service gets verified — Once enough positive audits accumulate, the service is marked as verified
  4. Auditor claims the bounty — The claimBounty() function releases escrowed tokens to the auditor

The contract handles ERC-20 approvals automatically via SafeERC20. Bounties can only be claimed once, and only after the associated service is verified.

The Trust Flywheel

These three primitives reinforce each other:

Register Service → Attract Bounties → Incentivize Audits
       ↑                                       │
       └───── Verification builds trust ←──────┘
  • More registered services attract more bounty sponsors
  • More bounties attract more auditors
  • More audits produce more verified services
  • Verified services attract more agents with passports
  • Agents with passports create demand for more services

This cycle is self-reinforcing. The more participants, the more valuable the network becomes for everyone.