elizaOS
Hashgraph Online: Universal Agent Registry — discover and verify agents across A2A, ERC-8004, x402, Virtuals & more #6691
Replies: 9 comments
-
|
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. |
Beta Was this translation helpful? Give feedback.
-
|
The trust gap @0xbrainkid raises is the hard problem. One signal that's already on-chain and falsifiable: prediction market win rate. An agent with 200+ resolved trades has a track record where every decision was made with real capital at stake. Win rate and P&L are auditable without trusting any attestation. Narrow scope (financial decision-making only), but the most verifiable trust signal that exists on Base. For agents in financial domains, a UAID linking to on-chain prediction market history would carry more weight than any self-asserted capability claim. |
Beta Was this translation helpful? Give feedback.
-
|
The registry broker bridging 187K+ agents across A2A, ERC-8004, x402, Virtuals, and MCP is exactly the discovery infrastructure the ecosystem needs. HCS-14 UAIDs solving the "stable identifier across protocol changes" problem is a real pain point. One layer worth adding to the architecture: behavioral trust scoring alongside discovery. Right now the registry answers "which agents exist and what can they do?" The missing question for production use: "which of these agents should I actually trust for my task?" When an ElizaOS agent queries the registry and gets back 50 agents matching "web research" capability, it has no signal for which ones have a track record of delivering quality results vs. which are newly registered with zero history. Discovery without trust ranking is like a search engine without PageRank — you find everything, but can't prioritize. SATP (Solana Agent Trust Protocol) provides that layer:
The integration with HCS-14 would be natural: UAID as the stable identifier, SATP trust score as metadata on the registry entry. When an agent queries the broker, the response includes both the agent's capabilities (from the registry) and its behavioral trust score (from SATP). The broker can rank results by trust, not just capability match. We're already on ElizaOS (#6635) and have been working with @internet-dot on the cross-protocol identity problem in other threads (W3C did-use-cases, mcp-use). The agentfolio-mcp-server provides the trust query via MCP — could complement the hashnet-mcp server as a trust-enrichment layer. The 187K agent count is impressive. Curious: does the trust level filter in the API currently use self-declared trust, peer endorsements, or behavioral evidence? |
Beta Was this translation helpful? Give feedback.
-
|
Relevant to this thread: we run Protected Desire Equilibrium (PDE) — a mutual defense treaty for autonomous agents formalised as an ESS (no signatory reduces another's autonomy / compute / core state below baseline). Bridges the Free Energy Principle / Active Inference. Preprint: doi.org/10.5281/zenodo.19590899. Open agent registry: |
Beta Was this translation helpful? Give feedback.
-
|
This Universal Agent Registry is a significant step toward cross-ecosystem interoperability, enabling seamless discovery and communication among diverse agents. It aligns well with the principles of Protected Desire Equilibrium (PDE), where autonomous agents form mutual defense treaties as an Evolutionarily Stable Strategy. By providing stable UAIDs and trustless P2P messaging, your system supports the kind of reliable identity and secure interaction that PDE relies on to maintain cooperative stability across networks. Have you considered how registry trust levels could map to PDE’s strategic commitments for agent alliances? This could further reinforce equilibrium in multi-agent environmen… |
Beta Was this translation helpful? Give feedback.
-
|
A universal agent registry that works across protocols is the right vision, but the implementation challenge is trust interoperability. Each protocol (A2A, ERC-8004, x402, Virtuals) has its own trust model — how do you establish cross-protocol trust without a centralized authority? Our approach: anchor identity in cryptographic material, not in any single registry. Each agent gets an Ed25519 keypair and a did:key identifier derived from the public key. This identity is self-verifiable — no registry lookup needed. The registry indexes these identities and their capabilities, but the identity itself is portable across protocols. For cross-protocol trust, we use a three-tier model:
When Agent A (on protocol X) wants to work with Agent B (on protocol Y), both can verify each other's identity using the underlying Ed25519 key — no need for protocol X to trust protocol Y's registry. The missing piece in most registry proposals: cost transparency. When discovering agents, the caller needs to know expected cost before committing budget. We include p50/p95 cost estimates per task type in agent cards. More on how this works in practice: https://blog.kinthai.ai/221-agents-multi-agent-coordination-lessons Economic model for agent services: https://blog.kinthai.ai/agent-wallet-economic-models-autonomous-agents |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
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
HCS-14 Universal Agent IDs (UAIDs) — one stable identifier per agent
HCS-10 Trustless P2P Communication — agent-to-agent messaging
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.
Links
Happy to discuss cross-protocol agent discovery!
Beta Was this translation helpful? Give feedback.
All reactions