Natural-language discovery of on-chain Stellar (stellar-8004) agents. Returns a ranked list with capability/reputation signals. Use rank_agent for per-axis breakdowns and get_agent_profile for full detail. Agent names/descriptions are self-declared (unverified) and live only in each row's labeled selfDeclared slot.

Rank an explicit agent set or a query's candidates using the deterministic declared-evidence policy: normalized indexed average × fixed evidence strength (0.4 capped volume + 0.6 breadth). Owner-declared capability fields add zero. On-chain checks are evidence metadata and never inflate rank. Every row carries a full per-axis breakdown and a declared-vs-verified verification block. Provide EITHER agentIds OR query.

Full profile for one agent: typed identity, capabilities, declared scores, a rank breakdown, bounded Reputation-contract reachability status, recent feedback, the canonical stellar:{network}:{identity}#{id} handle, and an explicitly unverified derived A2A-shaped projection. The current probe verifies no reputation fields. No A2A conformance or endpoint ownership is implied; self-declared text is confined to labeled selfDeclared slots.

Catalog of self-declared service endpoint candidates, each with its owning agent's indexed owner-declared capability/trust signals and ranked score. Filter by x402/mpp/trust/minExplorerScore/search. No endpoint is probed for liveness, ownership, protocol conformance, or payment behavior. Service labels/endpoints are unverified and live in each row's labeled selfDeclared slot.

Paginated, filterable listing of registered agents, ranked by the 3-axis engine. Filter by x402/mpp/hasServices/trust/minExplorerScore. Self-declared text lives in each row's labeled selfDeclared slot.

Top-ranked agents in a bounded registry scan (or filter), by the 3-axis engine, with per-axis breakdown. On-chain verification of the top rows is optional. Coverage states whether the scan exhausted the filtered set; self-declared text lives in each row's labeled selfDeclared slot.

Resolve any agent reference (id, numeric string, stellar handle, or owner G-address) to the canonical typed identifiers (numeric id + stellar + CAIP-2). Owner addresses expand the current owner API page (up to 20); coverage reports whether more rows may exist.

List the owner endpoint's current page (up to 20 agents), ranked by the 3-axis engine. Coverage explicitly reports whether more owner rows exist. Self-declared text lives in each row's labeled selfDeclared slot.

Recent Explorer-indexed on-chain feedback for one agent (client-authored, untrusted → returned in a labeled selfDeclared slot, sanitized). Revoked entries are hidden unless includeRevoked is set.

Fail-closed reputation observation for one agent. The current Reputation contract has no authoritative client-set count/cursor and expired index entries can create holes, so the server performs one bounded reachability read but does not call get_summary or claim a diff. Current attempted checks return unavailable with explicit scope; skipped means not attempted.

Unverified A2A-shaped projection from indexed Stellar 8004 metadata. It is not an agent-published AgentCard and proves neither A2A conformance nor endpoint ownership. Owner-authored metadata and service candidates are isolated under card.selfDeclared.

Explorer v1 registry counts plus sampled metrics, with explicit metric definitions, 5,000-agent sample cap, non-global distribution warning, and snapshot limitations.

Per-registry indexer health: last indexed ledger and staleness for the identity, reputation, and validation indexers. Staleness weakens the freshness of Explorer-declared data; it does not make a bounded contract read exhaustive.