Amazon added MCP Apps support to Amazon OpenSearch Service for interactive observability dashboards in an AI assistant
Amazon OpenSearch Service now supports MCP Apps – an extension to MCP in which the agent's response includes, in addition to text, an interactive visualization (trace waterfall, service topology, log pattern) directly in the AI assistant window, without the need to switch to a separate observability tool.
According to Amazon, Amazon OpenSearch Service now supports MCP Apps – an extension to Model Context Protocol that adds a so-called dual response pattern. When an AI agent calls a tool through MCP, the response contains two parts: a text summary for the agent's reasoning and an interactive visualization (a trace waterfall, service topology map or log pattern view), which is rendered directly in the AI assistant window alongside the text response.
The architecture includes three components: a local MCP server running on the user's computer, an agentic IDE (according to the article, for example, Claude, VS Code or Cursor) and the OpenSearch UI application, a serverless observability interface connected to OpenSearch domains, serverless collections, CloudWatch and Amazon Managed Service for Prometheus. The MCP server validates AWS credentials and forwards the query to OpenSearch UI, which executes it against the connected data sources and returns both a text summary and a visualization artifact that the IDE renders as an interactive widget. According to Amazon, both data and credentials remain in the customer's account because the MCP server runs locally.
The goal is to eliminate the so-called verification loop: the previous workflow required the user, after receiving a text hypothesis from the agent, to leave the IDE, log into a separate observability interface in the browser and manually repeat queries to verify the agent's conclusions. According to Amazon, the visualization generated by MCP Apps is deterministic because it is produced by running code against the same data that feeds standard dashboards, so the user is verifying the actual query result rather than the AI interpretation.
The source article also describes a specific example of trace investigation, but that section of the text was not available. You can find the details in the source article.
Why it matters
For engineers who use AI agents connected to Amazon OpenSearch Service when investigating incidents, the previous step of switching to a separate browser and manually repeating queries to verify the agent's conclusion is eliminated – the visualization appears directly in the conversation. This matters especially for organizations that, according to the article, deliberately chose a local agentic observability setup for control and cost reasons instead of a fully hosted vendor solution, because the verification loop was the main operational burden of this approach.
Two audiences, two different impacts
What this means
For individuals
Engineers investigating incidents through an AI agent connected to Amazon OpenSearch Service can now verify results (trace waterfall, service topology, log pattern) directly in the AI assistant window, without logging into a separate observability tool and rerunning queries.
For a business
According to Amazon, companies that run observability agents locally instead of using a fully hosted vendor solution have so far paid a price in the form of a manual verification loop outside the AI tool. MCP Apps move this step into the same chat thread, which may reduce the operational burden associated with incident investigation for teams using Amazon OpenSearch Service.
ProductivityCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.