Frequently asked questions
Answers for evaluation, security review, and procurement. Anything not covered here: [email protected].
Product
Does Lumière PayCheck hold or move our money?
No. We never hold, route, or escrow funds. Your agents pay sellers directly from your wallets; we only answer whether a payment should happen, with a signed receipt when it should.
What happens if Lumière PayCheck is unavailable?
We recommend agents fail closed: if they can't get an answer, they don't pay. If your wallet requires our receipts, payments pause until we're reachable. Live status and incident updates are at /status.
Which networks do you support?
Payment decisions work for any x402 network your sellers use; we check the network against what we observed for that endpoint, and you can restrict your workspace to specific networks. Our paid delivery tests and real-usage data cover USDC on Base and Solana, and you can pay us in USDC on either.
How accurate are the grades?
Grades come from free monitoring every 30 minutes and our own paid delivery tests (a failure only counts after a second paid check confirms it). A means we paid, it delivered cleanly, and it's up 95%+ of the time; B means it delivered with a caveat; C means it probably works but we couldn't confirm it (healthy endpoints we haven't paid-tested yet are C and caution until we do); D means it doesn't work as listed, but no money is lost; F means it cost money or is unsafe (avoid). Our own problems (rate limits, our network, a request the endpoint can't use) never count against a seller. Sellers can dispute a grade publicly, which triggers a paid re-test. Our tester pays up to $0.05 per call on its own; sellers of pricier endpoints can fund one test of their own endpoint at its own price (the money goes back to their payout wallet, the endpoint page says the test was seller-funded, and the grade comes only from what the test finds). Grades can't be bought. Grades are automated assessments, not guarantees.
Can we try it first?
Yes. The endpoint checks, API, and MCP server are free, and you can try Business free for 30 days with just an email (no card). If you don't continue, your agents keep the free allow/deny check; if you do, the same workspace and keys carry over. For Enterprise, we set up a custom trial workspace. Test keys let you integrate without affecting spending or reports.
Security and compliance
Where is data stored?
On our hosting provider (Render) in the US East (Virginia) region, with encrypted nightly backups stored off-site (Cloudflare R2). See the privacy policy for every provider involved.
What do you store about our agents and payments?
Workspace settings, hashed keys (never the keys themselves), and each payment decision's inputs and outcome for your audit period. We don't store IP addresses, and for key-less sign-in we store which workload acted, never its token.
Are you SOC 2 certified?
Not yet, and we haven't had a third-party penetration test yet. Our security page describes our controls in detail and lists current limitations openly. We're glad to complete your security questionnaire.
Do you offer SSO?
Yes, on Enterprise: sign in with Microsoft Entra ID, Okta, Google Workspace or any OpenID Connect provider, with your groups mapped to our roles, and SCIM 2.0 to add and remove people automatically. SAML isn't offered yet. On every plan, each person can also have their own key with a role, removable at any time, and agents can use key-less sign-in with your cloud identity.
Can an agent pay without asking PayCheck?
PayCheck never holds your funds, so an agent with its own wallet technically can. Two things cover that: our npm package never signs a payment PayCheck denied (and shows the agent as Enforced by SDK), and if you register the agent's wallet, every x402 payment it makes is matched to PayCheck's decisions, so a payment made without asking, after a deny, or for more than approved alerts you within the hour (Business) or 10 minutes (Enterprise). You can also have such an agent frozen automatically.
What do sellers learn about us?
Nothing, unless your Owner turns on agent credentials. Then sellers your agents pay learn only what you choose: that the agent belongs to a verified PayCheck workspace (anonymous), or also your company name and verified domain (named). Each seller sees a different agent ID. Limits, people and payment history are never shared, and turning it off works immediately.
How is an agent's risk decided?
From signals we can explain, over the last 7 days: payments outside PayCheck, spending spikes, denied requests, many new seller wallets, a new network range, and low-grade sellers. Every level comes with its reasons. It doesn't change anything unless you set a rule (send High-risk agents' payments to a person, or deny them).
Can our auditors verify the audit trail?
Yes. Decisions are hash-chained: any edit or deletion is detectable, and anyone with viewer access can run Verify audit trail. Each decision can also be replayed from its stored inputs.
Who can see our workspace at Lumière PayCheck?
Only for support, through a read-only view that expires after 15 minutes and is recorded in your activity log. Any change we make to your workspace (for example, custom limits you asked for) appears there too.
Commercial
How is Enterprise priced?
A flat annual or monthly fee, or usage-based: a base fee plus a price per authorization above an included amount, invoiced monthly through Stripe. See plans and pricing.
What's your uptime commitment?
We publish live component health and 90-day uptime at /status. A formal service-level agreement can be discussed as part of an Enterprise agreement.
What about contracts?
We'll work with your standard vendor process. Contact [email protected].
Can we export or delete our data?
Yes. Audit data exports as CSV (Business and Enterprise), and spending reports as CSV. Email us to delete your workspace.
Who operates Lumière PayCheck?
Lumière LLC, a Connecticut company.