Amazon OpenSearch Service has introduced MCP Apps, a new capability that integrates observability tools into developers' integrated development environments (IDEs). This innovation allows agents to provide both text-based summaries and interactive visualizations within the same conversation thread, eliminating the need to switch between tools for verification. The change aims to streamline the observability workflow by keeping the verification process within the agent loop, reducing manual effort and increasing efficiency.
MCP Apps extend the Model Context Protocol by adding a dual response pattern. When an AI agent calls an MCP App tool, it receives both a text summary and an interactive visualization rendered in the IDE. These visualizations are generated by executing queries against the same data sources that power dashboards, ensuring deterministic results. This approach allows developers to verify findings without leaving the IDE, improving the speed and accuracy of root cause analysis. The visualization includes elements like trace waterfalls, service topologies, and log pattern views, which are displayed alongside the text response, enabling real-time inspection and confirmation.
The source explains that the traditional observability workflow requires engineers to leave the agent loop to verify findings, which involves switching to separate observability tools and manually re-running queries. This process is time-consuming and disrupts the efficiency of agentic automation. With MCP Apps, the verification process remains within the IDE, maintaining the speed advantage of agentic workflows while preserving control over data and credentials. The solution addresses the primary operational burden of verification, which has historically been a bottleneck in local agentic setups.
Source: awsml
