Grafana Cloud Launches New Agent Observability Framework

Grafana Cloud just rolled out Agent Observability to help teams monitor, evaluate, and test their AI agents in production. AI agents are notoriously tricky because they can give different answers to the exact same question without ever throwing an error. They also turn a single user request into a long, messy chain of model and tool calls that is hard to debug. The new tool tracks latency, errors, token usage, and costs down to individual conversations. It also includes guards that use regex patterns and personal data detection to block bad behavior or redact sensitive information before the model sees it. Teams can even monitor coding assistants like GitHub Copilot and Cursor with a dedicated view that tracks token spend by user or repository. Free and Pro plans include thirty thousand generations and twenty-five million evaluation tokens each month.