利用 Bedrock AgentCore 在公用事业领域实施 MCP 以实现智能体 AI.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.I N D 3 8 9Operationalizing MCP for Agentic AI in Utilities with Amazon Bedrock AgentCoreKumar LakshmipathiGenerative AI Lead for EnergyAmruta KarnikPrincipal Soluti

2、ons Architect 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaMeet thy neighbor!Business Problem-Outage Management Setting a baseline-Agents,MCP,AgentCore Runtime&Gateway Whiteboard-Architecture Best PracticesNext Steps 2025,Amazon Web Services,Inc.or its affiliates.All righ

3、ts reserved.Outage Management&the Case for Agentic AIDynamic field conditions:Grid status,Asset availability.Complex coordination:Safety,Crews,Materials,PrioritiesSiloed data systems:SCADA readings,weather forecasts,outages,crew locations Real-time reasoning:Evaluate and adapt as conditions evolveOp

4、timization:Improve prediction,routing,and responseBusiness impact:Reduces downtime,costs,and customer impact 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Data Sources?SCADAGIS/asset registry Social MediaWeather feedsInventory systemsCrew management systems Customer information

5、systemsField mobility apps/IoT sensorsAgents?Detection AgentDispatch AgentCommunication AgentLogistics AgentAssessment AgentRestoration Agent 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.oProtocol standardizes how models discover&use tools&data.oRapid tool integration.No custom

6、 codeoConsistent governance&interoperability across toolsMCP Host(App/Agent/.)MCP ClientMCP ServerToolsResourcesPromptsStreamable HTTP ConnectionModel Context Protocol 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Agent RuntimeOperationalizing with Amazon Bedrock AgentCore Model