Pre-built Grafana dashboards for Gigapipe observability stacks.

Monitor any OTel-instrumented AI agent in a single dashboard — token usage, latency, cost estimates, and error rates across all your models and frameworks.

Panels included:

• Requests, error rate, total tokens, estimated hourly cost
• Token usage by model (input + output)
• Request latency — P50 / P95 / P99
• Request rate by model
• Error rate over time
• Input / output tokens by model (bar gauge)

Works with: Claude Code Agent SDK, Anthropic SDK, LangChain, LangGraph, OpenAI Agents SDK, GitHub Copilot SDK, and any framework that follows the OpenTelemetry Gen AI semantic conventions.

Requires: A Prometheus datasource pointing at your Gigapipe endpoint.

1. In Grafana: Dashboards → Import → Upload JSON file
2. Select ai-agent-observability.json
3. Set the datasource to your Gigapipe Prometheus endpoint

Full step-by-step setup — OTel Collector config, agent instrumentation for every major framework, and Grafana panel queries — in the blog post:

Your AI Agent Is Burning Money. You Just Can't See It Yet.