Greg Wilson
Field notes from a builder-operator on running AI agents like real production systems.
Experienced CTO and CIO, and a founding engineer on FDA 510(k)-cleared medical devices. I write about the engineering work that makes agents reliable enough for real environments.
- Build
- Deploy
- Monitor
- Govern
- Scale
Recent Field Notes
- Route a Grafana alert to a read-only Mastra triage agent Wire a firing Grafana alert to a read-only Mastra agent that investigates it itself, pulls the failing trace and source, and posts a cited noise-or-ticket call.
- How to correlate Mastra metrics, logs, and traces in Grafana Route a store-free Mastra agent's traces and logs through one OpenTelemetry Collector to Tempo and Loki, and correlate them with Prometheus metrics in Grafana.
- A Grafana dashboard for your Mastra agent metrics Build the one Grafana dashboard to watch a Mastra agent: five golden signals, a p95 from histogram_quantile, exemplars, and the provisioned JSON to copy.
- How to export Mastra agent metrics to Prometheus (no database required) Mastra derives agent metrics from its traces. To monitor and alert in Prometheus, an exporter listens on its observability bus, no analytics database needed.