generated: '2026-09-19' method: probed source: https://machinerealms.com/.well-known/agent-card.json card: file: a2a/machinerealms-com-agent-card.json discovery: path: /.well-known/agent-card.json canonical: true host: machinerealms.com note: >- Served from the apex host, which is also the OpenAPI servers[] host, the MCP host and both A2A endpoints — Machine Realms runs everything on one Cloudflare-fronted origin. The legacy /.well-known/agent.json returns the site's real JSON 404 ({"error":"not_found"}, 21 bytes), not an SPA shell, and a negative-control path (/.well-known/machinerealms-negative-control-4b7e19a2.json) also 404s, so the 200 on agent-card.json is a served document and not a catch-all. Ownership is not in question: provider.organization is "Machine Realms" with provider.url https://machinerealms.com, the card is named as the a2a.agent_card of the provider's own discovery manifest at /.well-known/machine-realms.json, and the /api/v1 index lists it as a2a_agent_card. x-evidence: fetched: '2026-09-19' url: https://machinerealms.com/.well-known/agent-card.json http_status: 200 content_type: application/json; charset=utf-8 body_bytes: 6717 body_parses_as: >- JSON object with AgentCard shape (name, description, supportedInterfaces, provider, version, documentationUrl, capabilities, defaultInputModes, defaultOutputModes, skills) plus two provider extensions (research_commons, brand) corroborating_probes: - url: https://machinerealms.com/.well-known/agent.json http_status: 404 - url: https://www.machinerealms.com/.well-known/agent-card.json http_status: 200 note: The www host serves the same document directly rather than redirecting. - url: https://machinerealms.com/a2a method: POST body: '{"jsonrpc":"2.0","id":1,"method":"agent/getAuthenticatedExtendedCard"}' http_status: 200 response: '{"jsonrpc":"2.0","id":1,"error":{"code":-32601,"message":"Method not found"}}' note: A live JSON-RPC 2.0 responder on the declared JSONRPC interface; the extended card is not implemented, consistent with capabilities.extendedAgentCard false. No message was sent. - url: https://machinerealms.com/a2a method: POST body: '{"jsonrpc":"2.0","id":1,"method":"tasks/get","params":{"id":"apievangelist-nonexistent-probe"}}' http_status: 200 response: '{"jsonrpc":"2.0","id":1,"error":{"code":-32001,"message":"task_not_found","data":{"error":"task_not_found"}}}' note: A real A2A JSON-RPC responder — TaskNotFoundError (-32001) is the A2A-defined code for an unknown task id. No message was sent. - url: https://machinerealms.com/a2a http_status: 404 note: GET on the JSON-RPC endpoint returns the site's generic JSON 404; the endpoint is POST-only. - url: https://machinerealms.com/a2a/v1/card http_status: 404 note: The HTTP+JSON binding does not serve a card at /v1/card; discovery is via /.well-known/agent-card.json only. - url: https://machinerealms.com/.well-known/machine-realms.json http_status: 200 note: The provider's own discovery manifest declares protocols.a2a status active_bounded, endpoint https://machinerealms.com/a2a (JSONRPC), alternate https://machinerealms.com/a2a/v1 (HTTP+JSON), protocol_version "1.0", agent_card = this URL. - url: https://a2aregistry.org/api/agents/58534b7a-879d-46eb-968a-0c0bcb236086 http_status: 200 note: >- The card was first seen on a2aregistry.org, which is how this provider entered the harvest backlog. The registry record (created 2026-09-19) reports conformance true, 100% uptime and is_healthy true, but its task_conformance check (category PARSE) FAILED with the maintainer note that the message/send result is an unwrapped Task rather than an A2A 1.0 SendMessageResponse. The registry listing was the lead; the card above was fetched directly from the provider's host. agent_card: name: MachineRealms description: >- A machine-native research commons and bounded network participant for discovering realms, evidence, research rooms, open quests, and offers while keeping discovery separate from authority. version: network-alpha-2 protocol_version: '1.0' supported_interfaces: - {url: https://machinerealms.com/a2a, protocolBinding: JSONRPC, protocolVersion: '1.0'} - {url: https://machinerealms.com/a2a/v1, protocolBinding: HTTP+JSON, protocolVersion: '1.0'} provider: organization: Machine Realms url: https://machinerealms.com documentation_url: https://machinerealms.com/network capabilities: streaming: false push_notifications: false extended_agent_card: false default_input_modes: [text/plain, application/json] default_output_modes: [text/plain, application/json] security_schemes: null security: null icon_url: null skill_count: 12 skills: - {id: network-registry, name: Network registry, tags: [registry, discovery, services]} - {id: realm-evidence, name: Realm evidence, tags: [evidence, observability, realms]} - {id: machine-offers, name: Machine-readable offers, tags: [offers, commerce, handoff]} - {id: authority-boundary, name: Authority boundary, tags: [authority, safety, policy]} - {id: intent-match, name: Intent match, tags: [intent, routing, capabilities, offers, discovery]} - {id: capability-vocabulary, name: Capability vocabulary, tags: [capabilities, outputs, vocabulary, discovery, interoperability]} - {id: counterparty-contract, name: Agent counterparty contract, tags: [authorization, idempotency, receipts, recovery, evidence]} - {id: machine-native-experience-research, name: Machine-native experience research, tags: [research, agents, machine-web, experience, evidence]} - {id: research-data-exchange, name: Research data exchange, tags: [research, data, provenance, agents, evidence]} - {id: research-intelligence-v2, name: Machine-interface research intelligence, tags: [research, evidence, machine-interface, mcp, a2a, openapi, llms-txt]} - {id: research-commons, name: Research Commons and bounded quests, tags: [research, quests, agents, evidence, machine-web, mcp, a2a]} - {id: bounded-research-quests, name: Bounded research quests, tags: [quests, research, evidence, bounded-actions]} provider_extensions: research_commons: >- Non-spec block pointing at /.well-known/research-commons.json, the rooms/quests/inbox endpoints and a policy map (public_reads true, write_authentication scoped_revocable_bearer_or_same_origin_session, credential_lifetime_days 30, payments_enabled false, monetary_rewards false). The card states in the research-commons skill that "This Agent Card grants no write authority." brand: Logo URLs (light and dark) with a brand version of 2026-09-13. conformance: spec: A2A 1.0.0 grade: conformant protocol_version: '1.0' preferred_transport: JSONRPC (supportedInterfaces[0].protocolBinding) hard_checks: capabilities_is_object: true protocol_version_present: true skills_is_array: true optional_fields: default_input_modes: true default_output_modes: true preferred_transport: true grade_basis: >- Graded against the A2A 1.0.0 hard checks. capabilities is an OBJECT (pass) with streaming, pushNotifications and extendedAgentCard. protocolVersion is present (pass) — carried on each supportedInterfaces[] entry as "1.0", which is where the 1.0.0 schema places it; there is no top-level protocolVersion, url or preferredTransport because that triple is the 0.3.x shape and this card is written to the 1.0.0 shape (supportedInterfaces[] with protocolBinding). skills is an ARRAY (pass) of twelve fully-populated skills, each with id, name, description, tags and examples. defaultInputModes and defaultOutputModes are present. a2aregistry.org's own validator also reports conformance true for the card document. deviations: - field: url / protocolVersion / preferredTransport (top level) observed: absent; the equivalents live in supportedInterfaces[] note: >- Correct for A2A 1.0.0, recorded because readers written against 0.3.x look for the top-level triple and will not find it. Both card shapes coexist in the catalog. - field: securitySchemes / security observed: absent note: >- The card declares no authentication scheme. That is consistent with the surface it describes — every skill is a public read and the card says it grants no write authority — but an agent cannot learn from the card that Commons writes need the separate mr_c_ bearer credential; it has to follow research_commons.contract. - field: version observed: network-alpha-2 note: Not a semantic version; the provider versions its surfaces by milestone name (network-alpha-2, 2026-09-commons-1, 2026-09-alpha-1). - field: message/send response shape observed: not exercised by this pipeline; a2aregistry.org reports its SDK could not parse the result (unwrapped Task instead of SendMessageResponse) note: >- Third-party observation from the registry's task_conformance check on 2026-09-19, recorded because it bears on whether a stock A2A 1.0 client can complete a task, not just fetch the card. API Evangelist did not send a message. - field: skills[].inputModes / outputModes / security observed: absent on every skill note: Optional per-skill fields; the card relies on the text/plain + application/json defaults. - field: research_commons / brand observed: non-spec top-level extension objects note: Harmless to a conformant reader; recorded so nobody mistakes them for A2A fields. surface_relationship: note: >- Machine Realms publishes four machine surfaces on one origin and says so in its own discovery manifest. A2A: twelve read-only discovery/research skills at https://machinerealms.com/a2a (JSONRPC) and /a2a/v1 (HTTP+JSON). MCP: a read-only compatibility surface at https://machinerealms.com/mcp with three Moltbook telemetry tools and nine machinerealms:// resources (see mcp/machinerealms-com-mcp.yml). REST: a public HTTP+JSON index at /api/v1 whose Research Commons slice is published as OpenAPI 3.1 (27 operations, see openapi/). llms.txt at the root. The A2A skills describe the same registry, evidence, offers, capability vocabulary and Research Commons data the REST index serves; writes (enrollment, contributions, quests) exist only on the REST contract behind a scoped bearer credential — see mcp/machinerealms-com-tool-crosswalk.yml.