B2B integration case study
How agent platforms, research desks, and fintech backends add market intelligence without API keys, monthly seats, or opaque vendor lock-in — using x402 USDC settlement and a free accuracy leaderboard for procurement due diligence.
Audience: Engineering leads, agent operators, and procurement evaluating pay-per-call intel APIs.
Base URL: https://conc-exe.xyz · Contract: OpenAPI 3.1 · Trust signal: GET /api/concierge-intel-accuracy
Problem
- Traditional market-data APIs require API keys, monthly minimums, and opaque usage billing.
- Autonomous agents need machine-readable discovery (OpenAPI, MCP) and predictable per-call pricing.
- Procurement teams lack observable trust signals — no public track record for desk-style verdicts.
Solution
Concierge ships 24 paid x402 routes plus free discovery and security surfaces (scope validation, Concierge self-audit). Agents pay only for calls they make via x402 (Solana or Base USDC, or TCX on supported routes). No signup required for consumers.
| Layer | What you get |
|---|---|
| Structured intel | Macro, wire, DeFi desks, alpha signals — JSON per function (Intel APIs) |
| Pay-per-call | Raw $0.02 · signal $0.10 · bundle $0.25 — see Pricing |
| Discovery | GET /openapi.json, GET /.well-known/api-catalog, GET /api/mcp |
| Trust | GET /api/concierge-intel-accuracy — free verdict track record vs 24h BTC |
Example workflow
A trading agent runs a morning desk brief before routing capital:
- Macro context —
POST /api/concierge-intel-macro($0.02 raw tier) - Wire headlines —
POST /api/concierge-intel-wire($0.02) - Desk verdict —
POST /api/concierge-intel-verdict($0.10 signal tier) - Trust check —
GET /api/concierge-intel-accuracy(free) before escalating to human review
# Free trust probe (no payment)
curl -s https://conc-exe.xyz/api/concierge-intel-accuracy | jq '.evaluated'
# Paid intel via x402 client (402 → sign → retry)
pay --sandbox curl https://conc-exe.xyz/api/concierge-intel-macro -d '{}'Use the Playground to inspect 402 challenges before wiring production agents.
Verdict accuracy methodology
Each settled intel-verdict call records the desk signal and BTC mark. After 24 hours, the system scores alignment:
| Signal | Hit condition (24h BTC move) |
|---|---|
snipe / follow | BTC ≥ +1% |
avoid | BTC ≤ −1% |
watch / rebalance | |move| < 2% |
The leaderboard exposes evaluated.hitRatePct, per-signal breakdowns, and recent snapshots. This is social proof for routing decisions, not a performance guarantee.
Live accuracy JSON → · documented in OpenAPI as getConciergeIntelAccuracy
Integration paths
OpenAPI
Build any HTTP client — x-payment-info, ApiError schema, examples on every operation.
MCP Registry
xyz.conc-exe/concierge-intel v1.1.0 — Cursor/Claude at /api/mcp.
Agent SDK
@conc-exe/agent — discover → pay → call.
AgentCash
npx agentcash add https://conc-exe.xyz — install all skills into MPP agents.
pay.sh
pay curl — sandbox wallet, handles 402 automatically.
Procurement checklist
| Question | Verify at |
|---|---|
| Is the API contract public? | GET /openapi.json |
| Are paid routes listed with prices? | GET /.well-known/x402 |
| Is there a machine index? | GET /.well-known/api-catalog |
| Can we audit verdict track record? | GET /api/concierge-intel-accuracy |
| Rate limits documented? | X-RateLimit-Limit headers · Agent readiness |
| MCP listed on official registry? | registry.modelcontextprotocol.io |
Agent card & discovery
The platform agent card at GET /.well-known/agent-card.json now includes discovery.intelAccuracy and trust fields for A2A routing. Per-agent cards (agt_…) expose the same accuracy URL under discovery.intelAccuracy.
curl -s https://conc-exe.xyz/.well-known/agent-card.json | jq '.trust, .discovery'Discover UI loads live accuracy stats alongside x402, MCP, and OpenAPI probes.
Concierge · Executive Lounge