Kakao Releases FAQ for AI Agent Competition Participants; Bonus Points for Using Kanana and playMCP

An FAQ summarizing key anticipated questions and answers has been released to help participants understand the AI agent competition co-hosted by Kakao and the Korean Institute of Information Scientists and Engineers. This competition focuses on implementing Agentic AI that autonomously perceives the environment and performs actions to achieve goals, going beyond simple chatbots.

Participants can receive bonus points by using 'Kanana 1.5', the Korean LLM released by Kakao, as the core brain of their agent, or by directly registering tools on 'playMCP', Kakao's open MCP platform, or calling existing tools. Using either of these two technologies qualifies for bonus points, and organically combining both can lead to higher evaluations. There are no restrictions on using open-source frameworks such as LangChain and LlamaIndex, or local LLMs.

During the judging process, originality in problem definition, technical completeness, applicability, and utilization of Kakao technologies are comprehensively evaluated. In particular, governance perspectives such as reliability, safety, hallucination control, and data privacy protection are emphasized to mitigate risks associated with autonomous agent behavior. Additionally, performance metrics such as task success rate, robustness, autonomy, and efficiency must be considered during demos and presentations.