Agentic observability with Amazon OpenSearch Service MCP Apps
TL;DR
Amazon OpenSearch Service now supports MCP Apps, which return interactive visualizations alongside your AI agent's text responses. Learn how a single, locally run MCP server lets your agent move from alert to trace to logs to root cause in one conversation, and how you can verify every step inline without leaving your IDE.
Nauti's Take
Interactive visualizations inside the agent's own answer solve a real problem: observability debugging otherwise breaks across chat, dashboard and logs, and that break is where incident time disappears. The catch is coupling, because a locally run MCP server for OpenSearch moves lock-in from the UI into the agent stack rather than removing it.
Teams already on OpenSearch gain clearly; teams still choosing a stack should check how portable these MCP apps really are.
Sources
