A resilient renewable energy community relies on a community-scale microgrid that integrates PV power, wind generation, battery energy storage systems (BESS), and controllable loads. Achieving reliable and cost-effective operation under uncertainty (renewable intermittency, grid price volatility, forecasting error, and extreme events) requires both predictive intelligence and closed-loop operational decision-making. This paper proposes an agentic AI microgrid management framework which is built upon three major layers: Strategic Planner, Operational Supervisor, and Executor. A Large Language Model (LLM) acts as the central orchestrator that coordinates specialized agents for data retrieval and forecasting, dispatch optimization, execution, and monitoring. The framework includes key agents such as the Home Agent, Central Agent (Orchestrator), Data Search and Forecast Agent, Decision Agent, Execution Agent, and Monitor & Review Agent. The framework follows an agentic pattern where the LLM is not only a parameter extractor but also a decision-maker across different stages of the workflow. In addition, the orchestrator can reason over system state, select tools based on the state, and coordinate multi-agent actions using hierarchical orchestration and tool calling, as well as multi-agent, tool-rich operational frameworks emphasizing standardized interfaces (e.g., MCP) and closed-loop execution. The framework will enable (i) adaptive, auditable dispatch decisions, (ii) human-in-the-loop policy review, and (iii) robust operation under uncertain renewable supply and grid conditions.