The M-POWER proposal presents an advanced framework that (i) enhances the OpenBEERS digital platform [1] by integrating a comprehensive case study of Martigny to support participatory decarbonization and urban-scale building energy simulations, and (ii) augments the platform with AI capabilities delivered through an LLM-based co-pilot.. The centerpiece of this innovation will be a secure Interaction Interface (derived through existing SDialog toolkit [2]) which will function as a governed, privacy-aware gateway connecting an autonomous (LLM-based) AI agent (co-pilot) to external simulation resources. To address the complexities of processing and querying massive, multidimensional energy data, M-POWER will support bi-modal (text, voice) and multilingual (English, German, French) human-machine interaction in natural language. A key security feature will be the separation between the AI-agent and data sources (crawled open-source data such as the Swiss Federal Building and Lodging Register [3], and pre-computed simulation indicators stored in a local database) to maximize grounding of retrieved information. The developed and integrated AI-agent will operate through standardized and logged function calls via open protocols (co-sharing raw data, or simulation results) back to the LLM in a structured, factual, and reusable way [4]. The developed M-POWER platform will deploy a bottom-up participatory approach, allowing citizens to use a natural language prompt to correct energy consumption values, or to refine physical, technical, or occupational stats regarding their homes. This collaborative dialogue will ensure that simulation models are accurately calibrated with real-world information provided directly by the community. Anonymized logging of interaction will enable refining existing functionality including user-input for possible social studies related to participatory decarbonization.