Operations | Monitoring | ITSM | DevOps | Cloud

session start tool call tool call cost outcome AI coding agents have quietly become part of most engineering teams. They open files, run commands, call tools, and push changes, often for hours at a time. Yet when you want to know what an agent actually did, the answer usually lives in one developer’s terminal history, and it disappears when that session closes.

If your Grape API calls error! inside an Active Record transaction, the client gets an error response and the database keeps the writes. No exception, no warning, no rollback. The request looks failed from the outside and succeeded on the inside. A fix merged into Grape on October 3 and is queued for Grape 4.1.0. It’s not released yet, so every current Grape version still behaves this way. Here’s why it happens, how to check whether you’re affected, and what changes when 4.1 ships.

The standard fix for an N+1 query in SQLAlchemy is to eager load the relationship with selectinload(). On SQLAlchemy 2.1.0 and 2.1.1, that fix can change your results. A regression in those two releases makes selectinload() ignore part of the join condition on some many-to-many relationships. The query gets faster, and it also returns related rows that the relationship was defined to exclude. It’s fixed in 2.1.2, released October 2.

FastAPI 0.142.0 ships with OpenTelemetry built in. Install fastapi, set one environment variable, and every request produces a trace, metrics, and error logs. No middleware, no instrumentation package, no code. That’s a big change for a framework most Python teams instrument by hand. It’s also easy to misread. The defaults record more than some teams expect in one place, and less than others assume in another. Here’s what you actually get.

Your AI coding agent can read your code, run your tests, and open a pull request. Until recently it could not see what that code does in production. MCP servers from monitoring vendors close that gap. Connect one, and Claude Code, Cursor, or Copilot can pull the error, the trace, and the slow query behind a bug report without you copying anything out of a dashboard. Most major monitoring vendors now offer an official MCP server. But they aren’t interchangeable. Some only expose errors.