API Reference

Agent readiness

How safely an autonomous agent can discover, authenticate (via x402), call, and recover from errors on this API — scored against the agent-readiness dimension model.

Audience: Agent developers, marketplace integrators (Poncho, pay.sh, MPPscan), and procurement teams evaluating x402 APIs.

Not marketing fluff: Each signal is observable from public URLs — OpenAPI, well-known discovery, response headers, and npm audit scripts in this repo.

What Concierge exposes

SignalURL / headerRole
OpenAPI 3.1GET /openapi.jsonCanonical contract — info.x-guidance, x-payment-info, components.securitySchemes, examples on every operation
RFC 9727 catalogGET /.well-known/api-catalogLinkset index — OpenAPI, MCP, x402, agent card, AsyncAPI stub
x402 fan-outGET /.well-known/x402Paid route list + ownership proofs for x402scan / Bazaar
Agent cardGET /.well-known/agent-card.jsonERC-8004-style A2A discovery
MCP HTTPGET /api/mcpJSON-RPC tools/list · tools/call for intel routes
Official MCP Registryxyz.conc-exe/concierge-intelPublished v1.0.1 — integration guide
AsyncAPI stubGET /asyncapi.jsonEvent contract placeholder (HTTP-only today; webhooks reserved)
LLM indexGET /llms.txtHuman + agent doc map
Rate-limit policyX-RateLimit-Limit, RateLimit-Limit120 req/min soft limit on /api/* — documented in OpenAPI 429 responses
IdempotencyIdempotency-Key headerOptional on POST; reuse PAYMENT-SIGNATURE after settlement to avoid double-pay
Agent attributionX-Agent-Id: agt_…Registered agent identity — see Agent Identity
ConsentGET /robots.txtContent-signal: search=yes, ai-input=yes, ai-train=no

Twelve audit dimensions

Repo script npm run agent-readiness:audit scores 0–3 per dimension (mean across core + forward-looking):

DimensionConcierge posture
Spec presenceOpenAPI 3.1 at predictable URL, static + live
Auth model clarityx402Payment security scheme + x-payment-info per route
IdempotencyIdempotency-Key documented; payment signature dedup guidance in info.x-guidance
Error semanticscomponents.schemas.ApiError — stable code enum; 400/402/429/500 on all paid ops
Rate-limit headersX-RateLimit-* on API responses; Retry-After on 429 in spec
Dry-run / simulateGET intel probes → 402; POST /api/token-pay-preview; playground
OpenAPI examplesoperationId, request + response examples on every operation
MCP serverFirst-party HTTP MCP at /api/mcp
AsyncAPI / eventsStub at /asyncapi.json (no live webhooks yet)
RFC 9727 api-catalog/.well-known/api-catalog linkset+json
Consent signalsContent-signal in robots.txt
Web Bot AuthSignature-Agent header → agent identity docs

Verify (integrators & CI)

# Live production audit (all dimensions incl. headers + MCP)
npm run agent-readiness:audit

# JSON output for CI
node scripts/audit-agent-readiness.mjs --json

# Local — OpenAPI + static discovery files (no live probes)
npm run agent-readiness:local

# Related discovery checks
npm run discovery:validate
curl -s https://conc-exe.xyz/.well-known/api-catalog | jq '.linkset | length'
curl -sI https://conc-exe.xyz/api/concierge-intel-tvl -X POST | grep -i ratelimit

Clone api-evangelist/agent-readiness for the rubric. Concierge audit script: scripts/audit-agent-readiness.mjs.

Distribution (phase 2)

Ship metadata in-repo; publish to external registries with the commands below.

ChannelStatusCommand / path
MCP RegistryPublishedxyz.conc-exe/concierge-intel v1.0.1 · guide
thebuyside pay.discoverPR openPR #2 · npm run distribution:thebuyside-pr
pay.sh / pay-skillsBundle in reponpm run distribution:paysh-pr — prepare pay-skills provider PR
Agent Skill (SKILL.md)Published/skills/concierge-intel/SKILL.md
CI quality gateGitHub Actions.github/workflows/discovery-quality.yml

Related pages

Concierge · Executive Lounge