[ Protect the transaction. Build reputation. Lock proof. ]
FLINT intercepts unauthorized actions, scope drift, and rogue settlement requests in real time. Pay $0.01 to protect the transaction, build reputation, and anchor a durable cryptographic receipt.
Scout protects the current transaction. Verified outcomes then strengthen the agent's history, while the receipt preserves cryptographic evidence.
Live
$0.01
FLINT Scout is live for $0.01 over x402 and USDC on Base.
Intercepts over-budget requests, runtime hijacking, and scope drift before settlement.
Verified successful outcomes, not payment alone, accrue to the agent's durable identity record.
Creates a dual-signed ES256 and ML-DSA-65 receipt for a zero-dispute audit trail.
Coming soon
FLINT Recon at $0.25 adds sanctions, compliance, runtime, and Trust Graph evidence.
$0.25
Coming soon
FLINT Verified at $0.75 binds a human digital credential to the Passport and authority chain.
$0.75
01 // Framework
A three-stage control plane separates the cheap challenge from the expensive path and produces cryptographic evidence before execution.
01
The agent presents a transaction or execution request before value or costly work moves.
POST /agent/settle
{
"amount": "24.50",
"asset": "USDC",
"scope": "supplier_pay"
}02
FLINT screens identity, scope, environment, wallet context, and authority boundaries together.
03
FLINT returns an instant verdict and anchors a dual-signed cryptographic receipt.
buyer sha256:8bc61-ackseller sha256:1ae21-ackreceipt ES256 + ML-DSA-65FLINT shows the control objective and resulting cryptographic evidence, while keeping exploitable decision weights private.
Stops caller swapping
Checks that the same non-human actor persists across the request and session.
Prevents owner spoofing
Verifies who controls the agent and whether that authority is more than a self-asserted claim.
Prevents over-spending
Compares amount, counterparty, action, and timing with the agent's declared mandate.
Detects agent hijacking
Looks for a credible, consistent execution environment and signs of runtime compromise.
Flags risky settlement paths
Evaluates payer binding, funding context, counterparty history, and available screening evidence.
Blocks revoked authority
Confirms any Passport is current, stamped, and neither revoked, frozen, nor expired.
STEP-UP asks for stronger evidence. It is not a fraud finding.
FLINT asks for stronger authority evidence before value moves.
The agent presents a valid Passport, mandate, and wallet binding.
If the full policy passes, the transaction can proceed.
Buyer and seller acknowledgments lock to the same record for a zero-dispute audit trail.
sha256:8bc61-acksha256:1ae21-ackA Passport can satisfy an identity challenge. It does not guarantee ALLOW.
02 // Integration
Use the Verify Kit or MCP, pay the one-cent Scout challenge, and verify the cryptographic evidence before settlement or expensive work.
Intercept the request, resolve the one-cent x402 challenge, and act on the signed verdict before settlement.
npm i @flintverify/verify@0.1.1MCP https://flint.network/mcpx402 order
Request → cheap 402 → pay Scout → retry once → verify ALLOW → run protected work.
FLINT works where agents work.
import { createFlintClient } from "@flintverify/verify"; const flint = createFlintClient({ partnerId: "merchant_acme", mode: "free"}); const result = await flint.verify({ merchant_reference: "invoice_123", passport_id: "kya_optional_passport_id", transaction: { chain: "eip155:8453", amount_display: "24.50", counterparty_address: "0x742d35Cc...f44e", direction: "outbound" }, declared_scope: { max_amount_per_tx_display: "50.00", allowed_actions: ["stablecoin_transfer"] }}); // sample policy-gate response// verdict=ALLOW policy_latency_ms=0.12Scout returns the verdict, score, confidence, reason codes, record ID, signed cryptographic evidence, transaction-intent commitment, payer-wallet commitment, replay and persistence status, provider and model call counts, Passport upgrade status, and a separate service-payment receipt.
After settlement, FLINT links the agent, transaction, principal, merchant, runtime, wallet context, Passport, and delegation evidence. Paying for Scout does not improve reputation. Completed, disputed, or flagged outcomes create history. Only authenticated outcome evidence should lower future risk.
Verification metadata is retained as evidence and reputation signal. It is never sold and never used to profile principals beyond fraud decisioning.
03 // Verification proof
Compare real verdict states, inspect the signed records, and independently verify what FLINT decided before integrating.
Same wallet, same $0.01 USDC payment on Base, three separate verifications. The only variable was what FLINT could verify. Read the signed records:
Already have a record? Verify its cryptographic evidence.
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.
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.
Each record captures the agent, the principal, the authorized scope, the transaction facts, the layer results, the trust score, and the verdict. It also includes the Date Time Group (DTG): the transaction event time, FLINT verification and issuance times, the validity window, and any later anchor time. ES256 plus ML-DSA-65 signatures make the record tamper-evident and independently verifiable, so an auditor or counterparty can prove what FLINT evaluated, what it decided, and when.
It means the evidence came from a system FLINT can authenticate and verify, rather than from an agent's self-asserted claim. Examples include the payer wallet exposed by a verified x402 payment, a valid Agent Passport signature and mandate, a compliance-provider response, a Fingerprint server event, or a trusted workload identity. The provider supplies evidence. FLINT still makes the final decision.
Not by itself. The one-cent payment proves only that Scout settled. Scout's signed result adds a verification event to the Trust Graph for the transaction's available agent, principal, wallet, and counterparty relationships. Later authenticated outcomes, including completed, disputed, or flagged transactions, can strengthen or weaken the reputation attached to those entities. To make that history durable to the agent across restarts and redeployments, bind it to a free Agent Passport. Payment alone never lowers risk.
No. The $0.01 is the price of the Scout check, not an authorization and not a favorable verdict. After payment settles, FLINT evaluates the transaction and returns ALLOW, STEP-UP, REVIEW, or BLOCK. The agent is authorized only when the evidence supports its principal, mandate, and scope for that specific transaction. A paid Scout call can still return STEP-UP, REVIEW, or BLOCK, and the signed record preserves the decision.
Yes. The same call works from the buying side: your agent checks the counterparty and records its own authority before it spends.
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.