Large language models (LLMs) enable natural language interaction with enterprise systems. However, their integration into Manufacturing Execution Systems (MES) requires secure access to production data, reliable execution of operational tasks, and measurable accuracy. This thesis presents KBS-Chat, a comprehensive LLM-driven MES interface for manufacturing environments, based on the Chat with MES (CWM) framework. The approach adapts the full CWM pipeline to a new PostgreSQL-based database with a unique schema and introduces a custom Streamlit-based graphical interface for interactive queries and operational commands. The open-source CWM pipeline was adapted to an MES database for automotive-parts manufacturing while preserving its end-to-end workflow: request rewriting for entity grounding, multi-step operation planning, iterative execution with intermediate results carried in the conversation context, and response generation grounded in execution traces.