Security Model

MoneyClaw is designed around user-authorized agent payments.

The goal is not to make payments invisible. The goal is to make real payments bounded, inspectable, and easier to control.

Core Principles

• prepaid by default
• bounded spend
• dashboard approval by default, with optional account-level agent auto-approval configured ahead of time
• visible wallet and payment-task history
• system-managed execution rather than card-as-product-object
• dedicated account inbox

What MoneyClaw Is For

MoneyClaw is for real purchases and payment flows requested or approved by the user. Some accounts may also enable narrowly scoped agent auto-approval ahead of time so their connected agent can keep moving without a second dashboard click.

What MoneyClaw Is Not For

MoneyClaw is not a tool for:

• bypassing issuer controls
• bypassing merchant verification
• bypassing fraud systems
• bypassing KYC, sanctions, or geographic restrictions
• fabricating billing identity or verification data

Safe Default Behavior

• inspect wallet and payment-task state before acting
• keep spending prepaid and task-scoped
• keep inbox access secondary to the main payment-task flow
• inspect final transaction state before retrying

Public Boundary

This repo intentionally publishes the trust and safety model.

Internal operations, infrastructure hardening, incident handling, and provider-specific defensive runbooks are intentionally not included here.