API and x402 sellers.
An agent wants your service. Verify it before you fulfill, and attach the signed record to the sale.
FLINT Verify
Sell to AI agents. Keep what they pay you.
One call before you fulfill tells you if the agent is authorized to spend, and signs the evidence that protects the sale. $0.01 per verification.
Have a record? Verify it here.
The verification framework
How every request gets a verdict.
Who it is for
Verification at the point of exchange.
Marketplaces and platforms.
Screen every agent transaction at machine speed without adding a human to the loop.
Agents themselves.
Before your agent pays a counterparty it has never met, 1 call checks who it is dealing with and records your agent's authority to spend.
Before the exchange
Verify before pay. Verify before serve.
Buyer side
Verify before pay
A buyer agent calls FLINT before it pays. The check evaluates whether the counterparty and transaction support proceeding before money moves, whether the buyer is dealing with another agent or a merchant.
Seller side
Verify before serve
A seller or merchant calls FLINT before it serves the order or releases the good. The check evaluates the buyer agent's authority and available signs of compromise before fulfillment.
How verification decides
Six evidence categories, evaluated together.
FLINT does not publish the weights, thresholds, or vendor signals behind a decision, because that would hand the same map to whoever is trying to defeat it. What it does publish is what gets evaluated, in plain language, so builders and counterparties know what to expect.
Identity
Which non-human actor is presenting this request, and is its presented identity consistent across the session.
Principal attestation
Who authorized or controls the agent, and whether that authority is verifiable rather than merely claimed.
Authorization scope
Whether the amount, counterparty, action, and timing of this request fit the authority that was declared for it.
Environment
Whether the agent is operating in a consistent, credible runtime context for the session it claims.
On-chain history
What the wallet and counterparty context show about provenance, funding, and screening outcomes.
Lifecycle
Whether any passport in play is stamped, current, and neither revoked, frozen, nor expired.
Your first verdict
Getting STEP-UP? That's normal.
Most agents get STEP-UP on their first verification. It doesn't mean your agent did something wrong. It means FLINT hasn't seen proof of who stands behind it yet. Your agent showed up without a bulletproof vest, like a traveler coming into a country without a passport. No one knows you yet. Nothing bad has happened.
The STEP-UP response includes a remediation menu with several ways to establish trust. The fastest is minting and claiming a free Agent Passport, and here's why it's the best one: the passport is where reputation lives. Every verified transaction accrues to your passport-agent pairing for the life of the pairing, so the trust you earn today is still working for you a year from now. Complete a cure once and every verification after that reflects it. Nothing to redo, nothing to repeat.
BLOCK is different. BLOCK means affirmative risk or fraud signals were present. If you got STEP-UP, you are not in that category.
What the counterparty sees
Both sides keep the same signed verification record.
The buyer side and seller side retain the same signed verification record as evidence of what FLINT checked and decided before money moved. When both sides acknowledge a completed transaction, the mutual acknowledgment becomes the shared receipt attached to the record they hold. If a dispute, a chargeback equivalent, or an examiner ever asks what was known at the moment of payment, the signed record is the answer, and it is tamper-evident: the hybrid signature (ES256 plus ML-DSA-65) proves the record was not altered after FLINT issued it.
Any record that carries an authorization scope also carries a scope evidence basis, so a reader can tell how that scope was established. passport_mandate means the scope is the agent's or its passport's own asserted authority; on its own, it cannot lift a verdict past STEP-UP. wallet_corroborated means the scope was independently confirmed against verified wallet or transaction evidence. A counterparty deciding how much weight to place on a record can read that field directly instead of guessing.
FLINT screens every agent, with or without a passport. A passport gives reputation a durable home across instance restarts, but it is never required for a verification decision.
Integration
Ship the check into your workflow.
Verify Kit
Put the check inside the agent's workflow.
The planned @flintnetwork/verify package will add verify-before-pay and verify-before-serve helpers. The package name is a release placeholder until publication. FLINT MCP tools are available now for agents that discover and call tools at runtime.
Integration quickstart
The 402 handshake, in 3 steps.
# 1. Call the paid endpoint with no payment attached
curl -i https://merchant.example/paid
# -> HTTP/1.1 402 Payment Required
# {
# "accepts": [{
# "scheme": "exact",
# "network": "base",
# "asset": "USDC",
# "payTo": "0x742d35Cc6634C0532925a3b844Bc454e4438f44e",
# "maxAmountRequired": "10000"
# }]
# }
# 2. Build and sign the payment for the offer, then retry with it attached
curl https://merchant.example/paid \
-H "X-PAYMENT: <base64-encoded-payment-payload>"
# 3. The seller verifies the paying agent with FLINT before fulfilling,
# and returns the resource with a signed verification record attached.Step 1
Call the endpoint
Your agent requests the paid resource with no payment attached, exactly like any other HTTP call.
Step 2
Read the 402, attach payment
The seller returns HTTP 402 with the accepted payment terms. Your client builds and signs the payment for those terms and retries the same request with it attached.
Step 3
Seller verifies, then fulfills
The seller checks the paying agent with FLINT before it releases the resource, and returns the signed verification record alongside the response.
The v2 x402 client packages, @x402/fetch and @x402/evm, wrap this handshake so an agent's own fetch calls handle the 402 retry and the EVM payment construction automatically.
What you get back
A decision you can act on and evidence you can keep.
Every call returns a verdict: ALLOW, STEP-UP, REVIEW, or BLOCK, with a trust score and the reasons behind it. Works on every agent, with or without a passport: identity signals, wallet provenance, runtime environment, and behavior across the network all count. Each decision ships as a signed, tamper-evident record you keep. When there is a dispute, a chargeback, or an examiner in the room, the evidence is already in your files.
ALLOWSTEP-UPREVIEWBLOCK
FLINT works where agents work
MCP connectorflint.network/mcp
Claude connectorClaude.ai, Claude Code, and Cowork
x402 Bazaar listingVerification endpoint listed
$0.01 per verification. Pay as you go over x402, USDC on Base, no account needed. Volume pricing when you need it: contact@flint.network.
Verification metadata is retained as evidence and reputation signal. It is never sold and never used to profile principals beyond fraud decisioning.
Proof and FAQ
See it verified. Get your questions answered.
Live proof
Same wallet, same $0.01 USDC payment on Base, 3 separate verifications. The only variable was what FLINT could verify. Read the signed records:
FAQ
FLINT Verify, answered.
My agent got STEP-UP on its first FLINT verification. Did I do something wrong?
No. STEP-UP is the normal first verdict for an agent FLINT hasn't seen before. Your agent has no passport, no claimed owner, and no declared scope, so there is nothing to verify yet. The response gives you several ways to establish trust; the free Agent Passport is the fastest and the most valuable, because once your agent has one, every verified transaction builds reputation on that passport-agent pairing for as long as it exists. You complete a cure once, and it carries forward to everything your agent does after. BLOCK is a different verdict reserved for affirmative risk or fraud signals.
Does my agent need a passport?
No. A passport is never required. It is where an agent's reputation durably accrues across instance restarts and redeploys. FLINT screens every agent, with or without a passport.
What is in the signed record?
Each record captures the agent, the principal, the authorized scope, the layer results, the trust score, and the verdict, signed with a hybrid scheme (ES256 plus ML-DSA-65) so it stays verifiable as quantum-capable adversaries emerge. Because it is signed, the record is tamper-evident evidence: you, an auditor, or a counterparty can later prove what was verified and what was decided.
Can I use Verify to protect my own agent's purchases?
Yes. The same call works from the buying side: your agent checks the counterparty and records its own authority before it spends.
Is FLINT a payment rail?
No. FLINT is rail-agnostic and holds no funds. It verifies the agent and returns a verdict; the rails move the money. We work alongside x402, Circle, Stripe, and the card networks rather than competing with them.