Enables AI agents to make secure, human-approved purchases on the web using local-first payment methods.
Sponsored
AgentPay provides a secure, local-first solution for AI agents to conduct online purchases without direct access to sensitive user financial data. It functions as an open-source Model Context Protocol (MCP) server, ensuring that credentials remain encrypted on the user's machine and all transactions require explicit human approval via cryptographic signatures. Payments are executed through a headless browser using a placeholder injection method, where real values are only briefly present in the DOM, further enhancing security and giving users full control over agent spending through a dedicated dashboard.
Key Features
01Local-first credential encryption and processing
02Budget and spending limits managed via a secure dashboard
032 GitHub stars
04Human approval for purchases with cryptographic signatures
05Open-source for full security auditability
06Placeholder injection for secure payment processing
Use Cases
01Allowing AI agents to autonomously purchase goods or services online
02Providing secure and auditable payment capabilities for agentic applications
03Integrating managed payment solutions into AI agent platforms like Claude Desktop or Cursor
