Grafana Labs has launched AI Observability in Grafana Cloud (public preview), a purpose-built solution for monitoring AI agents in production. Unlike traditional observability tools that track CPU, latency, and error rates, AI Observability captures agent-specific signals: conversations, tool calls, token usage, costs, and output quality evaluations. It is OpenTelemetry-compatible, instruments via a thin SDK, and integrates natively with Grafana Alerting. Teams can use LLM-as-a-judge, heuristics, or regex to detect bad outputs, drill into full conversation threads for debugging, and track agent versions automatically when prompts or toolsets change. The feature originated as an internal hackathon project built while developing Grafana Assistant, and is now available to all Grafana Cloud users.
Sort: