AWS engineers troubleshoot agentic AI with OpenTelemetry and OpenSearch
On July 7, 2026, AWS/OpenSearch engineers showed how OpenTelemetry traces and OpenSearch Agent Health can make agentic-AI failures inspectable before they hit production. The New Stack's walkthrough centers on troubleshooting agent workflows with OpenTelemetry and OpenSearch, while OpenSearch describes Agent Health as an open-source evaluation and observability framework for AI agents. For practitioners, the useful signal is a portable trace-plus-benchmark pattern for spotting wrong tool calls, latency and cost spirals, and regressions in agentic AI systems before a customer incident exposes them.
The useful takeaway is that agent observability is becoming a combined tracing, evaluation, and regression-testing problem, not just another logs dashboard. Agentic systems fail through tool choice, reasoning paths, latency loops, and cost cascades, so teams need telemetry that preserves the steps leading to an answer.
What happened
The New Stack published a July 7 walkthrough of AWS engineers troubleshooting agentic AI with OpenTelemetry and OpenSearch. The piece points to the OpenSearch Agent Health framework, which uses OpenTelemetry traces and OpenSearch-backed evaluation data to inspect agent behavior.
Technical context
OpenSearch's own Agent Health materials describe an open-source framework for observing and evaluating AI agents, including timeline views, benchmark runs, golden-path trajectory comparison, and local or OpenSearch-backed storage. AWS documentation also describes agentic AI features in Amazon OpenSearch Service, including chat, investigation, and memory features that operate within the user's existing permissions.
For practitioners
The operational pattern is portable: instrument agent steps with OpenTelemetry, store traces and benchmark results in a queryable system, and evaluate whether a prompt, tool, or model change moved behavior toward or away from the expected path. That is most useful for production agents that call tools, run recursive workflows, or need incident-ready explanations.
What to watch
Teams should watch whether OpenTelemetry GenAI conventions and OpenSearch Agent Health become practical defaults for agent QA, especially in CI pipelines and incident-response workflows. The harder test is whether these traces help teams prevent regressions, not merely explain failures after users report them.
Key Points
- 1AWS and OpenSearch engineers showed how OpenTelemetry traces can expose agent steps, tool calls, and failure paths.
- 2OpenSearch Agent Health combines trace observability with benchmark storage and golden-path evaluation for agent workflows.
- 3Practitioners can adapt the pattern for CI gates, production debugging, incident response, and cost-latency regression checks.
Scoring Rationale
This is a practical observability and tooling story for teams building agentic systems, not a major platform launch. It is useful because it connects OpenTelemetry traces, OpenSearch storage, and agent evaluation into an actionable debugging pattern.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
