generated: '2026-07-31'
method: searched
source:
- https://github.com/ards-project/ard-spec/blob/main/spec/ard.md
- https://agenticresourcediscovery.org/glossary/
- https://github.com/ards-project/ard-spec/blob/main/conformance/bin/conformance-test
type: Taxonomy
note: ARD's taxonomy is a media-type registry, not a subject taxonomy — the standard deliberately classifies
  by ARTIFACT KIND and delegates the meaning of each kind to that artifact's own protocol. Reproduced
  as the standard states it.
classification_axis: IANA media type on each catalog entry (`type`)
media_types:
- type: application/a2a-agent-card+json
  artifact: A2A Agent Card
  iana_status: de-facto community standard, formal registration pending
- type: application/mcp-server-card+json
  artifact: MCP server card
  iana_status: de-facto community standard, formal registration pending
  note: renamed from application/mcp-server+json by ADR-0008
- type: application/ai-catalog+json
  artifact: a nested catalog / bundle
- type: application/ai-registry+json
  artifact: a dynamic ARD search endpoint
  note: the entry type that carries a registry's base URL — how the API is discovered
- type: application/agent-card+json
  artifact: generic agent card
- type: application/agent-skills+zip
  artifact: packaged skill
- type: application/agent-skills+gzip
  artifact: packaged skill
- type: text/markdown; profile="urn:air:agent-skills"
  artifact: skill as markdown
- type: application/ai-registry
  artifact: registry (unsuffixed variant)
endpoints:
- path: POST /search
  requirement: MUST
  returns: entries ranked 0-100 by semantic relevance
  note: '`text` required, `filter` optional'
- path: POST /explore
  requirement: OPTIONAL
  returns: facet aggregations over the matched set
  note: 501 when unimplemented; does not federate
- path: GET /agents
  requirement: OPTIONAL
  returns: deterministic paged list
  note: ListResponse requires an `items` array
discovery_mechanisms:
- mechanism: well-known URI
  detail: https://{domain}/.well-known/ai-catalog.json
- mechanism: robots.txt directive
  detail: 'Agentmap: https://example.com/catalog.json'
- mechanism: HTML link tag
  detail: <link rel="ai-catalog" href="...">
- mechanism: DNS
  detail: SVCB with TXT fallback under _index._agents / _catalog._agents
  note: the docs and the spec disagree on the labels and record types
federation_modes:
- auto
- referrals
- none
ingestion_pipelines:
  required:
  - web crawl of ai-catalog.json
  optional:
  - git repositories
  - npm
  - OCI registries
trust_layer:
- identity (SPIFFE / DID / HTTPS FQDN)
- attestations
- provenance
- detached JWS signature
formal_schemas:
- format: CDDL (RFC 8610)
  role: authoritative abstract syntax
  file: json-schema/ard.cddl
- format: JSON Schema 2020-12
  role: runtime manifest validation
  file: json-schema/ai-catalog.schema.json
- format: OpenAPI 3.1.0
  role: the registry REST interface
  file: openapi/_original/ard-discovery-openapi.yml