generated: '2026-09-11' method: probed source: https://platform-docs.sarj.ai/.well-known/agent-card.json discovery: path: /.well-known/agent-card.json canonical: true host: platform-docs.sarj.ai notes: >- Found on the documentation host, not the API host. platform-api.sarj.ai, www.sarj.ai and sarj.ai all returned 404 for both /.well-known/agent-card.json and the legacy /.well-known/agent.json. Not an SPA catch-all false positive: the same docs host returns a text/plain 404 "Asset not found" for /.well-known/agent.json and /.well-known/ai-plugin.json, so the 200 is a genuinely served document. conformance: spec: A2A 1.0.0 grade: conformant protocol_version: '0.3' preferred_transport: HTTP+JSON hard_checks: capabilities_is_object: true protocol_version_present: true skills_is_array: true optional_fields: preferred_transport: true default_input_modes: true default_output_modes: true deviations: - id: supported-interfaces-instead-of-additional-interfaces detail: >- The card declares additional transports under `supportedInterfaces`; A2A 1.0.0 names this field `additionalInterfaces`. Passes every hard conformance check, but the field name is non-canonical, so a strict 1.0.0 parser will ignore the declared interface. - id: protocol-version-not-dotted-patch detail: >- `protocolVersion` is "0.3" rather than a full "0.3.0" triple. Accepted by most clients, but the A2A version string is specified as a semantic version. card: name: Sarj.ai description: Voice AI platform — build production AI-powered phone calls url: https://platform-docs.sarj.ai/ version: 1.0.0 provider: Sarj.ai documentation_url: https://platform-docs.sarj.ai/ capabilities: streaming: false push_notifications: false default_input_modes: - text/plain default_output_modes: - text/plain skills: 1 skill_ids: - sarj changed_since_last_pass: - field: skills[0].description detail: >- Rewritten to name call scheduling, retry policies and webhooks. Tracks the four operations that shipped to the REST contract; the card is being maintained. file: sarj-ai-developer-api-agent-card.json discovery_index: url: https://platform-docs.sarj.ai/.well-known/agent-skills/index.json http_status: 200 schema: https://schemas.agentskills.io/discovery/0.2.0/schema.json saved_to: well-known/sarj-ai-developer-api-agent-skills-index.json digest_published: sha256:856e9f515005ebfd4e5b9cc6e522f9b43aa64d057f9727ec289f97e24103b528 digest_verified: true note: >- NEW since the 2026-08 pass. An agentskills.io discovery document sits alongside the card and publishes a sha256 digest of the skill. We recomputed it over the file we harvested and it matches byte-for-byte, so the integrity claim is real and checkable — unusual on this surface. linked_agent_skill: url: https://platform-docs.sarj.ai/.well-known/agent-skills/sarj/skill.md http_status: 200 content_type: text/markdown; charset=utf-8 saved_to: skills/sarj-ai-developer-api-sarj.md bytes: 9696 sha256: 856e9f515005ebfd4e5b9cc6e522f9b43aa64d057f9727ec289f97e24103b528 note: >- The agent card's single skill points at a provider-published Agent Skill document. This is a genuine provider-authored skill, harvested verbatim rather than generated. Rewritten and expanded by the provider since 2026-08 (9532 -> 9696 bytes, substantially restructured): it now covers scheduled calls, cancel and reschedule, retry policies and permanent_recording_url, and it drops the speech-API material. x-evidence: fetched: '2026-09-11' url: https://platform-docs.sarj.ai/.well-known/agent-card.json http_status: 200 content_type: application/json negative_probes: - url: https://platform-api.sarj.ai/.well-known/agent-card.json http_status: 404 - url: https://platform-api.sarj.ai/.well-known/agent.json http_status: 404 - url: https://www.sarj.ai/.well-known/agent-card.json http_status: 404 - url: https://platform-docs.sarj.ai/.well-known/agent.json http_status: 404