Creating a transparent, sector-specific database of actions that would result in carbon emissions reduction is essential for guiding companies toward effective, data-driven pathways to meet their net-zero commitments. Information on carbon emissions reduction levers is scattered around greenhouse gas emissions disclosures and sustainability reports in dense text forms, and no systematic, sector and region specific reduction lever libraries are available to companies. This research proposes a multi-agent system leveraging Large Language Models (LLMs) integrated with Retrieval-Augmented Generation (RAG) to systematically extract, classify, and validate carbon reduction actions from publicly available sustainability reports. By constructing a standardized database of reduction levers categorized by industry, geography, and greenhouse gas scopes, this work empowers companies to prioritize high-impact, cost-effective emissions reduction strategies. We plan to integrate environmentally-extended input-output models to ensure that these actions are closely tied to sector-specific emissive sources, increasing their relevance and scalability. This initiative is expected to support companies in mitigating greenhouse gas emissions by offering a practical resource that accelerates the transition to a low-carbon economy, and makes actionable insights readily available to corporations, industry and the research community.