GitHub Copilot

GitHub Copilot

Monitoring dashboard for GitHub Copilot usage and performance via Application Insights OpenTelemetry data. Tracks agent operations, LLM token consumption, model latency (response duration and time-to-first-token), chat sessions, tool calls, feature adoption, and errors. Includes trace drill-down for debugging individual Copilot interactions.

Screenshots

Issues and feedback

https://github.com/1w2w3y/grafana-dashboards/issues

Features

• Summary
  • Agent Summary Statistics (total operations, tokens, avg response time)
  • Chat & Tool Summary (total calls, tokens, avg TTFT)
• Usage Over Time
  • Operations Over Time
  • Token Consumption Over Time by Model
• Model Performance
  • Model Usage Distribution
  • Response Duration by Model (Avg / P90)
  • Time to First Token by Model (P50 / P90)
• Latency Trends
  • Avg Response Duration Over Time by Model
  • Time to First Token (P50 / P90) Over Time
• Agent & Feature Activity
  • Copilot Feature Usage
  • Top Tool Calls
  • Sessions, Turns & Tool Calls Over Time
• Token Efficiency & Cost
  • Token Usage by Model (Table)
  • Input vs Output Tokens Over Time
• Errors & Reliability
  • Exceptions Over Time
  • Exception Details
• Trace Drill-Down
  • Select a trace by operation
  • Recent Operations table
  • Trace Visualization

How it works

• Grafana queries Application Insights via the Azure Monitor data source using Log Analytics (KQL) queries against OpenTelemetry data.
• Variables:
  • am_ds — Azure Monitor data source
  • sub — Azure subscription
  • rg — Resource group (resolved via Azure Resource Graph for microsoft.insights/components)
  • res — Application Insights resource
  • copilotTraceId — hidden variable for trace drill-down, populated from dependencies table filtered by service.name == "copilot-chat"
• Default time range is 7 days.

Requirements

• Grafana 11.6+ with the Azure Monitor data source configured with access to the subscription containing your Application Insights resource.
• GitHub Copilot telemetry data flowing into Application Insights via OpenTelemetry. See Ingest data into Application Insights via OpenTelemetry Collector for the end-to-end setup — running the OTel Collector with the Azure Monitor Exporter, the VS Code settings.json keys (github.copilot.chat.otel.enabled, otlpEndpoint, etc.), and KQL verification queries against the dependencies table (cloud_RoleName == "copilot-chat").

Change history

• 5/1/2026 Point data-ingestion link to the published Microsoft Learn guide (aka.ms/amg/dash-doc/otlp-appinsights)
• 3/23/2026 Initial version