Observing AI agents and MCP tool calls, end to end
Trace multi-step agent runs, tool failures and runaway loops with OpenTelemetry and Grafana. A runnable demo, five production scenarios, and the practices that keep agents observable.
Ravindra KumarA FIELD GUIDE TO BUILDING IN THE CLOUD
Less noise.
More know-how.
Practical guides, lessons learned, and useful resources for your cloud-native journey.
Notes from Ravindra KumarTHE LATEST FIELD NOTE
AI EngineeringObserving AI agents and MCP tool calls, end to end
Trace multi-step agent runs, tool failures and runaway loops with OpenTelemetry and Grafana. A runnable demo, five production scenarios, and the practices that keep agents observable.
Ravindra Kumar14 Nov 2026 · 17 min read
FROM THE NOTEBOOK
Worth your next read.
Ravindra Kumar
In a default Kubernetes cluster, any pod can reach your database
Kubernetes pod networking is flat by default. How NetworkPolicies lock it down, a default-deny pattern that's easy to reason about, and the selector trap everyone hits once.
Ravindra Kumar
The AI platform engineer on Google Cloud: what the job is, and the platform you build
What an AI platform engineer actually owns on Google Cloud in 2026: the front door, managed and self-hosted serving, pipelines and evals, data, guardrails, agent governance and cost. With a reference architecture you can download as draw.io.
Ravindra Kumar
Delete the cloud key in your CI pipeline. You don't need it anymore.
Workload identity federation swaps long-lived service account keys for short-lived tokens. How it works, and how to set it up on AWS, Google Cloud and Azure.
Ravindra KumarSMALL STEPS. REAL PROGRESS.