MCP = Observability + Code, a Real-life Example

Our bot is hitting an error. We can see it in the distributed trace. Here, see what happened when we noticed it: Austin fired up Claude Code (hooked up to Honeycomb with its MCP tool) and got it to find the error, fix it, deploy, and check that the fix worked. It got a little overconfident at first, but the ending is happy.

00:00 - Austin's AI-Powered Bot

00:30 - Fire up Claude Code

01:12 - If you Prompt it Better, it'll Do Stuff Better

01:28 - Please Look at the Code

01:52 - Great Observability There, Austin

02:48 - Go Look in Honeycomb, See if we Fixed It

02:57 - It's Wrong

03:28 - It's Right

IRL this took 22 minutes; the video speeds up the AI agent interactions and cuts out waiting.
This video includes Austin Parker, Jessica Kerr, and Ken Rimple.