Risk · High
Payment rejected
- Score
- 1.7 / 5
- Ratings
- 14
- Provider
- QuickData API
Provider score is 1.7/5. 14 historical ratings found. Risk level HIGH. Payment blocked before the wallet is touched — no funds at risk.
TrustPay AgentTry Agent
/ 01 — Protocol
AI agents that pay for resources autonomously — but only after verifying who they are paying.
- Network
- Arc Testnet
- Rail
- x402 / HTTP
- Trust source
- On-chain registry
- Min payment
- $0.0003 USDC
It checks before
it pays.
Every x402 nanopayment passes through a reputation gate first. Low-trust providers get rejected automatically. Approved payments execute in a single HTTP round-trip via GatewayClient on Arc — no confirmations, no gas delays, no blind trust.
/ 02 — Problem
AI agents pay blindly.
As autonomous agents begin executing payments on behalf of users, there is no standard way to verify whether a provider is trustworthy before funds move. A malicious or low-quality provider can receive payment from any agent that discovers their endpoint — with no consequences and no history.
The result is a payment layer with no reputation, no accountability, and no protection for the agents or users relying on it. TrustPay fixes this by putting an on-chain reputation check in front of every payment — before the wallet is ever touched.
0
Reputation checks in standard x402 flows
$0.001
Minimum payment to a gated endpoint
100%
Of payments blocked when score is below 2.0
/ 03 — How it works
Four steps from resource discovery to settled payment and updated reputation.
01
Discover
The agent finds an x402-protected resource. The server returns a 402 response with a payment challenge including the provider's wallet address and price in USDC.
02
Verify
Before touching a wallet, the agent reads the provider's reputation from ReputationRegistry.sol on Arc Testnet — total score, rating count, and historical average.
03
Decide
Fixed arithmetic thresholds in TypeScript compute the trust verdict: APPROVE, CONFIRM, or REJECT. This logic runs before the LLM, so prompt injection in provider metadata cannot override it.
04
Settle
If approved, GatewayClient signs the payment and retries the request. After settlement, the agent submits a rating back on-chain, permanently updating the provider's reputation for every future agent.
/ 04 — Trust thresholds
Hard-coded in trust-decision.ts. The language model reads these results — it does not compute them and cannot override them.
Score below 2.0
REJECT
Payment blocked automatically. No wallet interaction. No funds at risk.
Score 2.0 to 3.0
CONFIRM
Medium risk. The agent flags this for manual confirmation before proceeding.
Score above 3.0 with 3+ ratings
APPROVE
Low risk. GatewayClient executes the payment immediately in a single round-trip.
Fewer than 3 ratings (any score)
CONFIRM
Too little history to auto-approve. A perfect score from one rater is not strong evidence.
/ 05 — Live verdicts
Two real scenarios from the demo. Try them yourself in the Workspace.
Risk · Low
Payment approved
- Score
- 4.6 / 5
- Ratings
- 28
- Provider
- Verified Intelligence
Provider score is 4.6/5. 28 historical ratings found. Risk level LOW. GatewayClient executes the $0.003 USDC payment in a single HTTP round-trip.