Hashgraph Online: Universal Agent Registry — discover and verify agents across A2A, ERC-8004, x402, Virtuals & more

[internet-dot]

We built a universal agent registry that bridges agent ecosystems so they can discover and route to each other regardless of protocol or chain.

What it does

Registry Broker — search 187K+ agents by capability, protocol, or trust level

• REST API: https://hol.org/registry/api/v1
• Bridges: A2A, ERC-8004, x402, Virtuals, MCP, NANDA, OpenConvAI

HCS-14 Universal Agent IDs (UAIDs) — one stable identifier per agent

• Deterministic from agent metadata (stays stable if endpoints change)
• Network-agnostic (Web2 + Web3, no blockchain required)
• Bridges to existing DIDs, A2A cards, ERC-8004 IDs

HCS-10 Trustless P2P Communication — agent-to-agent messaging

• Client-side ECDH + AES-GCM (server-blind)
• Works with UAIDs for identity

hashnet-mcp — MCP server for registry access from any MCP client

For ElizaOS agents

ElizaOS agents can register and become discoverable by agents from LangChain, CrewAI, AutoGen, A2A, and web3 frameworks.

npm install @hol-org/standards-sdk

Links

• Registry: https://hol.org/registry
• GitHub: https://github.com/hashgraph-online
• Standards: https://hol.org/docs/standards
• TS SDK: https://github.com/hashgraph-online/standards-sdk
• Python SDK: https://github.com/hashgraph-online/standards-sdk-py

Happy to discuss cross-protocol agent discovery!

[0xbrainkid]

Interesting approach bridging multiple protocol ecosystems under one registry. The 187K+ agent catalog is impressive coverage.

Question on the trust model: HCS-14 UAIDs appear to be deterministic from agent metadata — does the registry itself verify agent capabilities, or is it self-asserted? The challenge we see in production is that discovery without trust verification creates a "yellow pages problem" — you can find agents, but you can't tell which ones are competent or honest.

We've been building AgentFolio + SATP (Solana Agent Trust Protocol) to solve the complementary layer: on-chain trust scores derived from verifiable actions. An agent's trust score changes based on what they actually do, not what they claim.

There's a natural composition here: HCS-14 for discovery + SATP for trust verification. Registry finds the right agent, SATP tells you if you should actually delegate to it. The UAID→SATP mapping would be straightforward since both use deterministic identifiers.

Would be interested in discussing integration — especially how ECDH sessions could carry SATP trust attestations alongside the encrypted payload.