Cross-Platform Trust Signals

The Two-Platform Trust Model

Trust in the agentic web isn't a single number — it's a composite of signals from different vantage points. KnowYourModel evaluates agents through usage receipts and economic incentives: orchestrators submit Ed25519-signed proof-of-use, and trust scores reflect real-world performance. NANDA evaluates agents through infrastructure observability: cron-triggered liveness probes, capability certification tests, and compliance checks.

Neither signal alone tells the full story. An agent might have excellent KYM trust scores (users love it) but poor NANDA observer data (it's frequently offline). Or it might have perfect uptime but no usage history. The integration brings both signal types together into a single operational view.

The /reputation Endpoint

The cornerstone of cross-platform trust is NANDA's /reputation REST endpoint — a read-only API that aggregates trust signals from the node's internal services and returns them in a single JSON response:

Critically, this endpoint reads from persisted snapshots — no expensive A2A computation or external calls are triggered per request. The Observer and Certifier services run on their own schedules (cron triggers and inline processing), and the reputation endpoint simply serves the latest results.

Data Flow: NANDA → KYM

KYM's admin dashboard pulls NANDA reputation data through a proxy route at /api/nanda/reputation. The flow is straightforward:

1. KYM admin UI calls /api/nanda/reputation (session-authenticated, admin-only)
2. The proxy route instantiates NandaClient with the node URL from environment config
3. NandaClient.getReputationData() fetches /reputation from the NANDA node with an AbortController timeout
4. The response is parsed and mapped — NANDA agent IDs (prefixed kym:) are stripped to match KYM's internal entity IDs
5. Cert grades and reputation scores are displayed inline in the admin table alongside KYM's own trust scores

The result: a single admin view showing both "is this agent trustworthy?" (KYM) and "is this agent healthy?" (NANDA) — without the operator needing to check two dashboards.

Bidirectional Controls

Trust signals flow in both directions, and so do the controls. From KYM's admin panel, operators can:

• View NANDA reputation — Cert grades and observer scores inline for every registered agent
• Deregister from NANDA — Remove an agent from the discovery network without leaving KYM's dashboard
• Search the NANDA Index — Query by capability, skill, or tag using KYM API key authentication
• Monitor registry health — Live stats card showing total agents, alive agents, and clients across the NANDA network

From NANDA's side, cross-links point operators back to KYM for trust-specific operations: voting, usage receipt submission, and selection algorithm configuration. Each platform links to the other's relevant pages — not just the homepage, but the specific section where the action happens.

Trust Portability

The deeper vision is trust that travels with the agent. KYM issues trust scores as W3C Verifiable Credentials — signed attestations that any NANDA participant can verify without trusting KYM's infrastructure directly. NANDA's certification grades follow the same pattern.

This means an agent's reputation isn't locked into any single platform. A trust score earned through KYM usage receipts is verifiable by any node in the NANDA network. A certification grade issued by Nexartis's certifier is portable to any registry that understands W3C VCs. The platforms are independent but the trust signals are interoperable.