Observability Engineering second edition out now! 27 net-new chapters written for today's observability challenges.Get your copy

How Observability Makes Better AI and Human Investigators

A hands-on, on-demand workshop from Honeycomb and AWS on how the same telemetry that helps engineers investigate distributed systems gives AI agents the context to reach application-level root cause — and how to make the agents you build observable in production.

Instructions

If you missed the event, you'll need access to:

  • An AWS account with CloudShell access and permissions for EKS, IAM, Bedrock, and AWS DevOps Agent
  • A free Honeycomb account — sign up at honeycomb.io/signup if you haven't. The free tier includes everything this workshop uses, including Honeycomb MCP.

Use this guide to watch the video and follow along.

In this workshop, you'll:

  • Deploy a real microservices application on Amazon EKS and connect it to Honeycomb with a single exporter change, no application code required
  • Investigate a real latency issue using traces, BubbleUp, and Canvas, the way you would in production
  • Watch an AWS DevOps Agent investigate the same issue first without telemetry access, then again after Honeycomb MCP is connected, and compare the two results
  • Build and instrument your own AI agent, in about 60 lines of code, and watch a complete Agent Timeline populate with its reasoning, tool calls, and token usage