Agentic observability with Amazon OpenSearch Service MCP Apps

Observability agents are fast at querying alerts, correlating logs with traces, and generating root cause hypotheses. However, verification traditionally requires manual navigation through separate dashboards, which can slow down incident response. Amazon OpenSearch Service introduces MCP Apps, a feature that embeds interactive visualizations directly into AI agent responses, closing this verification gap.

MCP Apps extend the Model Context Protocol to include a dual response pattern: a structured text summary and an interactive visualization rendered within the chat window. This setup involves a local MCP server running on the user's machine, which acts as a secure bridge between the AI IDE and the OpenSearch UI application. When an agent calls an observability tool, the MCP server forwards the request to OpenSearch UI, executes the query against connected data sources, and returns both a deterministic visualization and a structured summary.

This architecture ensures that users can review detailed traces, service maps, or logs inline, without switching tools or re-running queries externally. The visualizations are generated server-side, ensuring they accurately reflect the underlying data. This setup supports a range of investigation tools, from alert triage to detailed trace analysis, all within a single conversation thread.

The integration aims to improve operational efficiency, especially for on-call engineers, by enabling faster resolution times through inline verification. The MCP server setup is straightforward, requiring installation and configuration within supported IDEs, and it maintains full control over data and credentials.

Overall, MCP Apps enhance agentic observability by embedding deterministic, interactive visualizations into AI responses, reducing context-switching, and streamlining incident investigations.