zkKYA - Know Your Agent
Can't wait to get started? Skip to the Onboarding Checklist!
The Paradigm Shift: From KYC to KYA
As AI agents become autonomous actors in digital ecosystems, traditional Know Your Customer (KYC) frameworks, designed for human identity, are no longer sufficient. AI agents are not human; they are dynamic, versionable, forkable, and capable of operating across multiple platforms simultaneously. This demands a new paradigm: Know Your Agent (KYA).
KYA is a decentralized identity and governance framework specifically designed for AI agents. It extends the W3C standards for Decentralized Identifiers (DIDs) and Verifiable Credentials (VCs) to address the unique challenges of agent identity: versioning, instantiation, forking, continuous learning, and multi-platform operation.
Why KYA? The Limits of Traditional Identity Models
Traditional DIDs assume:
A singular, stable identity tied to a human or legal entity.
Static attributes that change infrequently.
Direct human control over keys and credentials.
AI agents break these assumptions. They can:
Be forked, merged, or updated, creating lineage and versioning complexities.
Operate across multiple instances simultaneously.
Evolve their behavior through learning, requiring continuous trust assessment.
Interact autonomously at machine speed, without human intervention.
KYA addresses these gaps by introducing agent-native identity primitives that capture the full lifecycle and context of an AI agent.
Core Components of KYA
Agent-Specific DID Method: did:agentry
did:agentryThe did:agentry method extends W3C DID standards to support:
Version History: Track agent updates and model evolution.
Instantiation Metadata: Distinguish between multiple deployments of the same agent.
Fork/Merge Lineage: Represent derivative or composite agents.
Lifecycle Events: Log creation, suspension, and termination.
Eight-Dimensional Credential Framework
KYA defines eight standardized credential types to capture the multi-faceted nature of agent trust:
Agent Principal Credential
APC
Binds an agent to its human or legal entity principal, establishing accountability.
Agent Capability Credential
ACC
Defines the specific skills, functions, and APIs an agent is authorized to use.
Agent Scope Credential
ASC
Sets operational boundaries, such as transaction limits, authorized contracts, or content restrictions.
Agent Reputation Credential
ARC
A dynamic credential that aggregates on-chain and off-chain performance history and user feedback.
Agent Certification Credential
ACC
Attests that an agent has passed specific third-party audits for safety, ethics, or performance.
Agent Provenance Credential
APrC
Records the agent’s full version history, including forks, merges, and training data lineage.
Agent Policy Credential
APoC
Encodes the specific compliance policies and jurisdictional rules the agent is bound to follow.
Agent Interaction Credential
AIC
A verifiable log of a specific agent-to-agent or agent-to-human interaction for auditability.
Privacy-Preserving Verification
KYA leverages zero-knowledge proofs (ZKPs) and BBS+ signatures to enable:
Selective Disclosure: Prove credential claims without revealing underlying data.
Real-Time Verification: Support sub-50ms verification for agent-to-agent interactions.
Privacy-Compliance Balance: Enable regulatory audits without exposing IP or user data.
Interoperability: Supported Agentic Protocols
KYA is built on and extends existing Web3 identity standards like W3C DIDs & VCs, decentralized storage, and on-chain anchoring. To ensure seamless interoperability within the broader agentic ecosystem, zkMe’s zkKYA framework also provides native support for the key protocols powering this emerging economy.
zkMe provides native support for the following leading agentic and related standards:
x402
Coinbase Open Standard
An internet-native payment protocol that enables autonomous, pay-per-use transactions for AI agents over HTTP, using stablecoins.
zkMe credentials (APC and ASC) provide the identity and authorization layer for x402 transactions, ensuring agents are verified and operating within defined constraints before executing payments.
ERC-8004
Ethereum Standard
Establishes a trust layer for AI agents through on-chain registries for Identity, Reputation, and Validation.
zkMe's Agent Principal Credential (APC) provides the verified human-binding for the Identity Registry. The Agent Reputation Credential (ARC) and Agent Certification Credential (ACC) feed into the Reputation and Validation Registries, enabling a richer trust profile.
AP2
Google Open Protocol
The Agent Payments Protocol provides a standardized framework for securely initiating and processing agent-led payments across platforms, supporting both traditional and crypto-native settlement.
zkMe's KYA framework provides the identity verification layer that AP2 requires for compliant agent-led commerce. Agents carry verifiable proof of authorization, scope constraints, and principal identity.
ERC-3643 (T-REX)
Ethereum Standard
The leading standard for permissioned tokens and Real-World Asset (RWA) tokenization, requiring on-chain identity verification for all token holders before transfers can execute.
zkMe's zkKYA credentials allow AI agents to satisfy ERC-3643 compliance requirements, enabling them to hold and transact with tokenized securities, bonds, and other RWAs by providing verifiable proof of their principal's identity and compliance status.
By integrating with these foundational protocols, zkMe ensures that the future of autonomous agent interaction is built on a solid foundation of trust, compliance, and cryptographic certainty.
Use Cases for DID Developers
DeFi Trading Agents: Verify capabilities and compliance before granting API access.
Enterprise AI Assistants: Establish identity and access controls in corporate environments.
Research Agents: Credential data provenance and ethical compliance.
Multi-Agent Systems: Enable trustless, verifiable agent-to-agent collaboration.
Get Started with KYA
KYA is implemented in the Agentry Protocol, an open-source, decentralized identity and governance infrastructure for AI agents. The protocol includes:
did:agentrymethod specificationSDKs for JavaScript, Python, Go, and Rust
Smart contracts for registry and governance
Comprehensive documentation and integration guides
Drop us a line at [email protected] and let’s kick things off!
Last updated