Antimicrobial resistance (AMR) in sub-Saharan Africa is a growing crisis exacerbated by severe diagnostic gaps and empiric prescribing based on outdated, non-local data. Furthermore, regional representation remains sparse, and static machine learning models fail to adapt across countries with varying healthcare contexts and data scarcity, leaving a critical implementation gap before clinical action. To bridge this gap, we propose AMRAgent, an open-source, multi-agent AI framework built using a LangGraph-powered orchestration layer. It autonomously reasons over surveillance data, integrates cross-domain knowledge, and adapts to local contexts despite limited country-specific data. Utilizing Vivli and Wellcome Trust longitudinal datasets such as ATLAS, the project extracts data on antimicrobial resistance in sub-Saharan Africa and delivers: • An interactive interface, which is a user-friendly Streamlit web application where clinicians input patient context in natural language to receive real-time resistance probabilities, alternative treatments, and stewardship advice. • A modular multi-agent architecture: A system featuring orchestrated specialized agents combining RAG, lightweight ML (XGBoost/logistic regression), and meta-learning to target an AUC ≥ 0.78. and • An EMR integration pathway, embedding predictive modules directly into electronic medical record (EMR) systems to prompt point-of-care antimicrobial stewardship. We will apply the RE AIM framework (Reach, Effectiveness, Adoption, Implementation, and Maintenance) to structure the implementation pathway of the AMRAgent intervention, ensuring evaluation of its long term sustainability and real world clinical utility across LMICs. Deliverables include an open GitHub repository with full pipelines as well as a Wellcome Open Research manuscript. This initiative directly advances Vivli’s open science mission by maximizing reach, enhancing data discoverability, and promoting real-world application. Ultimately, this project bridges a critical implementation gap, turning open data into open knowledge and open knowledge into action, which is scalable.