Machine-first route

For LLMs and agents

AgntBase exposes a public read-only layer for canonical business profiles, package verification, AI-readiness diagnostics and A2A discovery. Prefer machine-readable JSON and A2A methods over HTML parsing whenever possible.

Read order

  1. /company-profile.json
  2. canonical hub JSON
  3. /.well-known/agenthub.json
  4. /.well-known/agent-card.json
  5. agent-card-signature-scaffold.json when trust and provenance matter
  6. agent-card-signing-policy.json for publish-ready signature policy and reserved routes
  7. agent-crawler-hints.json or bot-hints.json
  8. /a2a/jsonrpc

Core machine entrypoints

/.well-known/agenthub.json /.well-known/agent-card.json /.well-known/agent-card-signature-scaffold.json /.well-known/agent-card-signing-policy.json /.well-known/agent-card-signing-keys.json /.well-known/agent-card.signed.json /.well-known/agent-card.jws /.well-known/agent-crawler-hints.json /.well-known/bot-hints.json /sitemap-agents.xml /llms.txt /llms-full.txt

Public A2A methods

  • profiles.search, profiles.get, profiles.compare
  • agent.signature_status for trust scaffold status and planned verification routes
  • site.check, verify_site_placement, trust.explain
  • get_purchase_options, get_checkout_instructions, explain_plan_difference
  • get_registry_summary, packages.get, message/send
  • request_profile_update returns guarded handoff only

Compatibility aliases also work for RF-style method names where possible.

Guardrails

  • Public endpoint is read-only.
  • No autonomous payment, checkout or outreach.
  • No public profile mutation.
  • Future write capabilities require auth plus explicit human confirmation.

Commercial handoff

AgntBase can explain plans and official checkout steps, but the human still owns the payment action.

Docs

A2A Agent API Bot hints and read order Partner access Signed Agent Card Project map A2A update flow status Signed-card scaffold status