The agentic internet needs a
trust layer.
AI agents will negotiate, transact, and collaborate across the open web. But there's no standard way for them to verify each other.
Hyfa is a zero-knowledge verification protocol that makes agent identity cryptographically provable—without centralizing trust or exposing private data.
Become a Founding PartnerThe Problem
Discovery is evolving.
Verification isn't.
Agent discovery and connection standards are emerging—NANDA, Hashgraph HCS, Google A2A, Anthropic MCP. They answer: how do I find and connect to an agent?
They don't answer: how do I know this agent is who it claims to be?
| Approach | Problem |
|---|---|
| Trust the registry | Registries can be compromised or become gatekeepers |
| Centralized verification | Rebuilds the chokepoints we're trying to escape |
| Every agent verifies everything | Doesn't scale |
The Insight
Separate discovery from verification.
Discovery
Needs to be fast and searchable.
It doesn't need to be trusted.
Verification
Needs to be cryptographic and tamper-proof.
It doesn't need to be centralized.
When you separate these concerns, discovery can be federated and untrusted—because a malicious registry can't forge cryptographic proofs. Bad candidates simply fail verification.
What We Built
Cryptographic verification
beneath agent discovery.
DISCOVERY (Untrusted)
NANDA, Hashgraph HCS,
DHT networks
Fast, federated, searchable
VERIFICATION (Trustless)
Merkle proofs, ZK identity,
on-chain anchoring
Cryptographic, tamper-proof
Three capabilities that matter:
Complete trust chain
Not just agents—Hyfa verifies organizations, members, and agents. OrgFacts establish organizational identity. MemberFacts verify individuals. AgentFacts describe capabilities. The full chain is cryptographically anchored.
NANDA-native, ecosystem-bridged
We extend NANDA's structure and generate valid credentials for Hashgraph HCS-11, W3C DID, and future standards. Register once, verified everywhere.
Three-layer privacy
Every attribute can be public (readable by anyone), committed (ZK-provable without revealing), or encrypted (selectively shared with specific parties).
Privacy & Proofs
Verify claims without exposing data.
Hyfa implements a complete zero-knowledge proof system for agent data:
Identity proofs
Prove relationships without revealing secrets
- •"This agent belongs to a verified organization" (without revealing which one)
- •"This member is authorized to manage this agent"
- •"This organization is in the network" (Merkle membership)
Attribute proofs
Prove properties without revealing values
- •Equality: "My jurisdiction is US"
- •Membership: "I have HIPAA compliance"
- •Threshold: "Our employee count ≥ 100"
- •Range: "Response time is between 10ms and 100ms"
Selective disclosure
Share data progressively as trust builds
- •Start with ZK proofs (prove without revealing)
- •Share encrypted field keys with specific parties
- •Reveal plaintext when appropriate
Encrypted data exchange
Sensitive attributes are encrypted per-field. Organizations control which parties can decrypt which fields. Different agents can see different subsets of the same document. This extends to full encrypted data sharing between agents.
How It Works
Three layers,
clear trust boundaries.
| Layer | Question | Trust |
|---|---|---|
| Attestation | Is this entity valid? | Cryptographic (Merkle tree + on-chain root) |
| Discovery | Find entities matching X | Untrusted (federated indexes) |
| Metadata | What can this entity do? | Signed (OrgFacts / MemberFacts / AgentFacts) |
Discovery returns candidates. Hyfa verifies them.
Compromised indexes can't forge proofs.
What Exists Today
Working code, not a whitepaper.
Complete demo flow:
The full trust chain—from domain ownership to agent identity to attribute proofs—is implemented and working.
Why We Need Partners
Distributed trust requires
distributed infrastructure.
Hyfa works. But one organization running all the nodes isn't distributed trust—it's a database.
We need:
Identity verification partners
KYC/KYB providers, domain registrars, certificate authorities
Node operators
Infrastructure partners to run DHT nodes and verification services
Registry integrators
Teams building on NANDA, Hashgraph, or MCP who want cryptographic verification
Research collaborators
Institutions working on agent identity and decentralized systems
The Alternative
What happens without a trust layer?
Centralization
Whoever runs the biggest registry becomes the gatekeeper.
Fragmentation
Every ecosystem builds its own verification. Agents can't work across boundaries.
Fraud
Without verification, impersonation is trivial. Legitimate providers can't differentiate.
Founding Partners
Help build the trust layer
for the agentic internet.
What founding partners do:
- Shape protocol development and governance
- Operate infrastructure (nodes, indexes, verification services)
- Integrate Hyfa verification into their platforms
- Fund continued research and development
What founding partners get:
- Direct input on architectural decisions
- Early access to protocol development
- Recognition as infrastructure contributors
Contributing to standards being built by:
Want to learn more?
Have questions? Want to get involved?
We'd love to hear from you.