AI software company Konan Technology said on Tuesday it has launched a new model of its Konan LLM, combining its proprietary search engine with MCP tools to specialise in implementing agentic AI.

The company said the new Konan LLM model strengthens tool use capabilities and built-in search functions, aiming to implement AI agents capable of autonomous execution in enterprise environments.

The new model builds on the context understanding and reasoning capabilities of Konan LLM ENT-11, an integrated reasoning model, the company said. It said this enhances the model’s ability as an agent to accurately understand users’ instructions and carry them out.

It added that the model includes an NL2SQL (Natural Language to SQL) feature that accesses databases and performs analysis through natural language, enabling non-experts to take part in data-driven decision-making.

Konan Technology said it plans to improve the usability of an agent builder that allows users to design agents directly. It also plans to introduce a HITL (Human-In-The-Loop) function that combines human judgement at key decision points during agent execution to increase agent reliability.

Won-cheol Do (도원철), a managing director in the LLM group at the AI Research Institute, said, "We focused on designing the LLM to serve as a 'brain' that performs complex tasks beyond simple question-and-answer." He added, "We will drive practical productivity innovation as an agent that can be used immediately in a rapidly changing enterprise environment."