GitHub Copilot App Supports OpenTelemetry for Agent Monitoring

On September 22, 2026, GitHub announced OpenTelemetry support in the GitHub Copilot app. OpenTelemetry is an open-source framework for collecting observability data.

Enterprise administrators can export information about agent sessions, model requests and tool interactions to compatible monitoring systems. This can help teams investigate unexpected behavior and manage operations centrally.

OBSERVING AGENT EXECUTION: AI agents may make several model requests and tool calls before completing a task. The final output does not always explain delays or failures. OpenTelemetry support helps operators inspect the sequence of work.

WHAT OPENTELEMETRY PROVIDES: OpenTelemetry is an open-source approach to collecting observability information. Exporting agent activity to existing monitoring platforms may let teams investigate AI workflows alongside other applications.

SESSION-LEVEL CONTEXT: Related operations within one task are easier to understand when they can be viewed together. This can help distinguish a failed tool invocation from a model response problem.

MODEL REQUEST PATTERNS: Request counts and timing can reveal repetition or processing bottlenecks. The precise fields available depend on the implementation and collection settings, so administrators should verify what is actually exported.

TOOL INTERACTIONS: Agents can fail because external tools return errors or lack permission. A reasonable model-generated plan does not guarantee successful execution. Tool telemetry can support operational troubleshooting.

PRIVACY DEFAULTS: GitHub says prompt and response bodies are not exported by default. This reduces unnecessary collection of conversational content, but metadata can still reveal operational details. Collection settings deserve careful review.

ACCESS AND RETENTION: Session identifiers, timestamps and tool activity may become sensitive when combined. Organizations should decide who can inspect the data, how long it is stored and whether an external monitoring provider is appropriate.

USEFUL ALERTS: Telemetry becomes actionable when teams define meaningful thresholds for failures, delays or unexpected activity. Alert fatigue is a risk, so notifications should reflect the importance of the underlying problem.

OBSERVABILITY IS NOT VALIDATION: Knowing that a tool ran does not establish that the resulting code is correct. Testing and review are still needed to evaluate output quality. Execution monitoring and product quality answer different questions.

A PRACTICAL ROLLOUT: Administrators can test data collection in a controlled environment, verify which fields are present and establish access and retention policies before wider use. Selecting a few useful metrics can reduce maintenance effort.

OPERATIONAL OWNERSHIP: Shared monitoring may help development and operations teams investigate agent behavior using a common system. The data will have limited value without assigned responsibility for analysis and follow-up.

WHAT COMES NEXT: As agents perform more complex work, organizations will need to explain what happened during execution. OpenTelemetry is a step toward that visibility, not a guarantee of security or correct results.

Prompt and response content is excluded by default, according to GitHub. Organizations should still review collection settings, access controls and data retention before enabling monitoring.