AWS brings interactive proof into the agent loop with OpenSearch MCP Apps

Key points

• Amazon OpenSearch Service now supports MCP Apps that return interactive visualizations alongside an agent’s text response. Sources: N1
• The local MCP server authenticates with AWS credentials while queries and rendered results remain tied to the customer’s configured data sources. Sources: N1

Verification is becoming part of the interface

An observability agent can propose a root cause quickly, but an engineer still needs to inspect the underlying trace, logs, metrics, and service topology. MCP Apps bring those artifacts into the same thread so the claim and the evidence can be reviewed together.

Amazon describes the rendered query result as deterministic because OpenSearch executes it against the connected data rather than asking the model to invent a chart.

Sources: N1

Agent infrastructure is moving beyond model access

AWS is competing on the operating system around agents: permissions, gateways, observability, verification, and the cloud infrastructure that keeps those workflows running.

The earlier Trainium and governance story still matters, but this release shows a more immediate enterprise advantage—making agent output easier to audit during real operational work.

Sources: N1, S1, S2

Why it matters

Enterprise agents will be limited by trust and verification as much as by reasoning quality. AWS can turn its existing observability estate into an advantage if customers can inspect what an agent did without leaving the workflow.

Sources: N1

Sources

1. Agentic observability with Amazon OpenSearch Service MCP Apps — Amazon Web Services · 2026-08-25
2. Govern AI agent tool access with Amazon Bedrock AgentCore Gateway — Amazon Web Services · 2026-08-21
3. Amazon.com announces fourth-quarter results — Amazon · 2026-02-05