Monveil is built around a simple idea: users should understand what they are approving before authorizing a stablecoin payment.
Payment requests can contain inconsistencies that are difficult for users to evaluate quickly. An amount might differ from the original invoice, a merchant signature might be invalid, or a request might be expired or duplicated. Monveil's Guardian engine evaluates these conditions through deterministic verification checks.
We integrated Kimi through the Moonshot API to make the results easier to understand. Rather than asking an AI model to decide whether a payment is safe, Monveil uses structured verification results as the basis for plain-language explanations.
For example, in our tampering demonstration, a legitimate ₹2,000 invoice is changed to request ₹2,200. The verification layer identifies the amount mismatch, while the AI explanation is intended to help the user understand why the request needs attention.
This separation is important. Kimi supports explanation and usability, while deterministic checks remain authoritative. AI output cannot override a failed verification result or authorize a payment on the user's behalf.
During development, we also implemented a rule-based fallback so the application can continue presenting verification information when the external AI service is unavailable. At the time of writing, the Kimi API is returning a quota or insufficient-balance error, so fallback responses must not be confused with live Kimi-generated output.
Our goal with Monveil is to make payment verification more understandable without treating AI as a substitute for security controls.
Project: Monveil
Live application: https://monveil.vercel.app/
Source code: https://github.com/pallaviXD/monveil
Top comments (0)