Terms used across the Agentic Resource Discovery (ARD) specification and this site.
Any external capability an AI client can call on to perform a task: an agent, MCP server, Skill, API, workflow, AI Catalog entry, or similar. The umbrella term ARD is built around.
A product: one discovery service built on ARD — github.com/agentfinder. It is not the specification itself; many discovery services, named or not, can implement ARD.
A separate manifest standard that predates ARD and shares much of its vocabulary. ARD defines its own ARD entry and its own manifest at ard.json; the two are independent specifications.
An orchestrator, agent, or harness — a chatbot, IDE assistant, or automation — that discovers agentic resources through ARD and invokes the ones it selects over their native mechanisms. Also just client.
The unit ARD is defined over: the description of a single agentic resource in a form that can be found by search. Every ARD entry is a well-formed catalog entry, but not every catalog entry is an ARD entry — an entry without representative queries cannot be found by search.
The open protocol defined by this specification: how agentic resources describe themselves, how clients ask for them, and how discovery services answer — all before invocation.
A verifiable claim attached to a catalog entry's trustManifest (e.g. SOC2, HIPAA, GDPR) that a registry or client can fetch and check.
The document a publisher hosts at https://<domain>/.well-known/ard.json to advertise its agentic resources: a JSON document with an entries array of ARD entries.
A single agentic resource description within a manifest: its identifier, type, location (url or inline data), and optional metadata.
A service that indexes agentic resources and answers search queries (POST /search). The specification's formal term is Agent Registry (or registry); this site uses "discovery service" and "registry" interchangeably. Agent Finder is one such product.
How discovery services draw on each other. A search's federation parameter is auto (the service merges upstream results), referrals (it returns other services for the client to query itself), or none.
Examples of the execution mechanisms and resource types ARD wraps: Model Context Protocol servers, Agent-to-Agent agent cards, and Skills. ARD discovers them; each is invoked over its own mechanism.
The IANA media type in a catalog entry's type field, identifying what an artifact is — e.g. application/mcp-server+json, application/a2a-agent-card+json, application/ai-skill. ARD acts as an envelope over these.
Whoever hosts an ard.json describing one or more agentic resources, typically at https://<domain>/.well-known/ard.json.
An entry in a search response pointing to another discovery service the client may query itself (used with federation: "referrals").
The specification's term for a discovery service that exposes the standard REST search interface. See Discovery service.
The score (0–100) on a search result: a semantic relevance ranking from the discovery service. It is not a trust, compliance, or safety rating.
Sample natural-language queries (representativeQueries, 2–5 per entry) that registries use to build semantic search ranking.
Optional verifiable identity and trust metadata on a catalog entry (trustManifest): a cryptographic identity (e.g. did:web, SPIFFE), attestations, provenance, and a detached signature. ARD communicates these signals; it does not itself confer trust.
A catalog entry's domain-anchored logical identifier, urn:air:<publisher>:<namespace>:<agent-name>, where <publisher> is a verifiable domain. It roots identity in DNS and stays stable as infrastructure changes.
The conventional manifest location, https://<domain>/.well-known/ard.json, where discovery services look for a publisher's entries.