Integrating AI Agents with KakaoPay Payment Open API: Developing the MCP Agent Toolkit

As AI agents became able to manipulate external systems through function calling, integration with payment systems emerged as a core challenge. KakaoPay introduced Model Context Protocol (MCP) to address the framework dependency and scalability limitations of existing approaches such as OpenAI function calling or LangChain tool.

MCP is a standardized communication protocol between AI models and external tools. Through this, payment features can be used regardless of framework across various LLMs such as Claude and GPT, and host environments such as Claude Desktop and Cursor.

KakaoPay designed and provides a total of 8 Tools to support payment and subscription payment processes.

• Payment-related: demo payment, payment preparation, approval, status inquiry, cancellation
• Subscription payment-related: subscription payment request, status inquiry, deactivation

The overall architecture is structured so that an AI Agent calls the KakaoPay API through the MCP Server (Agent Toolkit), helping developers implement payment features in a standardized way across various AI environments.