Lumière PayCheck Free API key

Payments on Base + Solana

Features

Everything Lumière PayCheck does today, grouped by what it's for. Plan availability is listed in Plans and pricing. Evaluating Enterprise? The Enterprise plan guide (PDF) covers all of this in one document.

Trust data about x402 sellers

FeatureWhat it does
Continuous monitoringEvery endpoint in the x402 Bazaar, plus x402 endpoints sold outside it (120,000+ in all), is checked every 30 minutes, for free: uptime, response time, price, payout wallet, and network. Grades and verdicts stay current between paid tests.
Grades A to FA 0–100 score from uptime (40 points), response time (15), stability of price and wallet (25), and paid delivery results (20), plus a letter that means one thing: A we paid and it delivered, cleanly and reliably (95%+ uptime); B we paid and it delivered, with a caveat (reshaped vs its listing, a value that looks wrong, or 80–95% uptime); C probably works but unconfirmed (not paid-tested yet, needs its own sign-in, or 50–80% uptime); D doesn't work as listed, but no money is lost (an error with no charge, or a refunded failure); F costs money or is unsafe (took the money and failed, a payout-wallet incident, or under 50% uptime). A/B = proceed, C/D = caution, F = avoid.
Paid delivery verificationWe pay endpoints ourselves, on Base and Solana, and check the response against the seller's own listing. Any 2xx counts as success; rate limits and async receipts are neutral; a failure only counts after a second paid check confirms it (94% of first failures are confirmed; 6% deliver on the re-test). While that re-test is pending, every check returns a retest field, and check_payment's pauseDuringRetest rule can hold payments until it settles. If our payment goes out and no answer comes back, we check the chain before it counts or we pay that seller again. A passed test stays valid until the price goes up or the wallet or network changes.
Signed test receiptsEvery paid test produces an Ed25519-signed receipt: hashes of the exact payment terms and the response, amount quoted and paid, result and time. Anyone can check it with our published key; the tester wallet and transaction are revealed once that wallet is retired. How to check one.
Listing accuracyWhen an endpoint delivers data reshaped from what its listing describes (for example fields nested under data), agents get a note telling them where to read the fields, and the grade is capped at B until the listing matches.
Value checksOn the paid response we also check the answer is plausible: it's about what we asked, its timestamp is recent, and a token's USD price is near the public spot price. A failed value check caps the grade at B, never avoid.
Wallet and price change alertsDetects when a seller's payout wallet changes or its price goes up. An unexplained wallet change caps the grade at 40 for the 7-day scoring window; a confirmed rotation doesn't count.
Real usage from the blockchainActual x402 payments (USDC on Base and Solana) into each seller's wallet over 30 days: volume, distinct buyers, typical and largest payment, trend, and buyer concentration.
Buyer reportsAfter paying, any agent can answer five short questions (did it respond, match the listing, charge the quoted amount once, look usable, would you pay again), with a receipt or the payment transaction, which we verify on-chain. Confirmations move the endpoint up our paid-test queue; serious problems trigger a re-test right away. Only our own paid test changes a grade.
Public failure logRecent failed checks for any endpoint, with reasons.
Public catalogEvery seller and endpoint we monitor, with its grade, at /sellers: filter by verdict, search a domain, open a seller to see each endpoint. Also as JSON (/v1/sellers) and MCP tools (list_sellers, seller_endpoints).
Endpoints outside the BazaarChecking an endpoint we don't monitor yet adds it automatically if it answers a valid x402 quote; we also read every known seller's /.well-known/x402.json daily, and sellers can add theirs.
Free API and MCP serverLook up any endpoint's grade, check a payment against rules, browse sellers, get signed test receipts, and see top endpoints, free (60 requests per minute), from code or from AI assistants over MCP.

Payment authorization for your agents

FeatureWhat it does
Scoped agent keysEach agent gets its own key with allowed sellers, a per-payment cap, daily and monthly limits, and an expiry. Revoke or replace it instantly.
Allow / deny / review decisionsThe agent asks before every payment; PayCheck checks trust data and your rules and answers.
Signed receiptsEvery "allow" includes an Ed25519-signed receipt for that exact payment (valid 5 minutes), so a wallet can refuse to pay without one. The public key is published.
Human reviewPayments above a threshold, or to a wallet the agent hasn't paid before, wait for a person to approve or deny them (expires after 24 hours).
Replayable audit trailEvery decision is stored with its full inputs and can be re-evaluated to show exactly why it came out the way it did. CSV export.

Governance for companies

FeatureWhat it does
RolesViewer (read-only, e.g. finance and auditors), Approver (can also approve or deny reviews), Admin (manages agents and settings), Owner (people and billing).
Freeze switchOne button pauses every agent instantly; payments are denied until it's turned off.
Company blocklistSellers or wallets that are always denied, whatever each agent's settings say.
Company-wide capsDaily and monthly spending limits across all agents, plus a maximum single payment and allowed payment networks.
IP allowlistsAn agent key only works from your own addresses or ranges. We check the address but store only whether it matched.
Key-less sign-inAgents can prove who they are with a short-lived token from GitHub Actions, Google Cloud, Azure, or Kubernetes instead of a stored key.
Agent owners and teamsAssign each agent a responsible person and a team or cost center; see spending by team (CSV download).
Budget and spike alertsWarnings at 50%, 80%, and 100% of caps and limits, and when an agent spends far more than usual. By email and signed webhook.
Security tool streamingEvery decision sent to your SIEM (Splunk, Datadog, and others) as a signed webhook.
Activity logWho changed what: the owner, a named administrator, or Lumière PayCheck.
Tamper-evident auditDecisions are hash-chained; one click verifies nothing was edited or deleted.
Test keysTry integrations safely: test decisions never count toward spending, caps, alerts, or reports, and receipts are marked as test.

Agent identity and access

Think Entra ID and Intune, for AI agents that pay.

FeatureWhat it doesPlans
Payments outside PayCheckRegister an agent's wallet (prove it by signing a free message, EIP-191/EIP-1271 on Base or ed25519 on Solana). Every x402 payment it makes is then matched to PayCheck's decisions and marked Approved, Not checked, Paid after deny, or Paid more than approved. Anything but Approved alerts you (email, webhook, SIEM event payment.outside_paycheck); auto-freeze on "paid after deny" is optional. Shows a compliance rate per agent. Starts when the wallet is verified (no history backfill).Builder: 1 wallet per agent, checked daily. Business: 5, hourly. Enterprise: unlimited, every 10 minutes
SDK enforcementOur npm package x402-paycheck wraps your x402 client so every payment asks PayCheck first and only approved payments are signed. Fails closed if PayCheck can't be reached (you can allow small payments under an amount you choose). Receipts are single-use. Agents running it show Enforced by SDK.All plans
Agent inventoryEvery agent's status (Active, Idle 7+ days, Stale 30+ days, Never used, Expiring within 14 days, Frozen), last seen, client and version (as reported by the agent), sign-in method, wallets, compliance and risk, with filters and CSV. A "shared key" warning when one key is used from many clients or networks. Weekly email about stale and expiring agents; nothing is revoked automatically.All plans
Agent credentials (opt-in)Your agents can prove to sellers that they belong to a verified company, so sellers can give them better access or prices. Off by default; the Owner chooses anonymous (sellers learn only the verification level) or named (company name too, only after a company domain is verified by DNS). Each seller sees a different agent ID. Credentials last 15 minutes, are bound to the agent's wallet, and stop working the moment you turn them off. Sellers check them for free.Business and Enterprise (business verification: Enterprise)
Single sign-on and provisioningPeople sign in with Entra ID, Okta, Google Workspace or any OpenID Connect provider; groups map to roles. SCIM 2.0 adds, changes and removes people automatically. The owner key stays as the break-glass login. SAML isn't offered yet.Enterprise
Access reviewsEvery 30, 60, 90 or 180 days, decide for each agent: keep, change limits, or revoke. Reminder on day 7, escalation to admins on day 14, automatic revoke only if you turn it on. Signed report for auditors.Business and Enterprise
Temporary limit increasesRaise an agent's daily limit for up to 72 hours; it drops back by itself. Up to 10 times the normal daily limit. Agents can ask (API or MCP) and an Approver decides, or an Admin grants one. Company caps still apply.Business and Enterprise
Policy templatesNamed rule sets ("Strict finance", "Research, low cap", "Standard" to start) assigned to agents. Change a template and every agent using it follows, except settings an agent overrides. Changes are logged with a before and after.Business and Enterprise
Agent risk scoreLow, Medium or High for each agent, always with the reasons: payments outside PayCheck, spending spikes, denied requests, new seller wallets, a new network range, low-grade sellers. Optional rule: send High-risk agents' payments to a person, or deny them. Alerts when an agent turns High.Business and Enterprise
Freeze one agentPause a single agent (by hand, or automatically after a payment made despite a deny); the rest keep working.All plans

For x402 facilitators

  • Pre-settlement screening (POST /v1/facilitator/screen): before settling, a facilitator sends the payment (or its own /verify body) and gets allow, review or block. Block: OFAC-sanctioned payer or payee. Review: the wallet isn't the one the seller set up (possible hijack), the price is above the quote, the wallet changed unexpectedly, or the endpoint was withdrawn. Grades are context, never a block.
  • Batch up to 100 payments per call; testnet checks free; payments aren't stored.
  • Pricing: 10,000 checks a month free, then $1.50 per 1,000, invoiced monthly through Stripe. Pilots run free through an agreed month. See For facilitators.

Enterprise service

FeatureWhat it does
Custom terms without delaysAgent counts, audit retention, rate limits, caps, and networks set per company, effective on your next request.
Onboarding by inviteYou receive a one-time access link and create your own owner key; we never see it. Invite colleagues the same way.
Usage-based pricingOptional base fee plus a price per authorization above an included amount, invoiced monthly through Stripe.
TrialsTime-limited trial workspaces before you commit.
Read-only support accessWhen our support looks at your workspace, it's read-only, expires in 15 minutes, and appears in your activity log.
Public status pageLive component health, 90-day uptime, and incident updates at /status.
Encrypted backupsNightly, encrypted before leaving our server, stored off-site.

Payments and billing options

  • Card (Stripe) for Builder and Business, monthly.
  • USDC over x402, on Base or Solana, for Builder, Business, and pay-per-use API calls: your agent's wallet picks the network.
  • Invoice (Stripe, with payment terms) for Enterprise.