GitHub has updated its Copilot usage metrics API to include detailed activity from agent applications. This enhancement allows teams to track the usage of agents, such as those from Claude and Codex, which can run directly within GitHub workflows.
Key Points
- The Copilot usage metrics API now reports agent app activity, broken out by individual agent.
- This activity is available in enterprise, organization, enterprise-user, and organization-user 1-day and 28-day reports.
- A new optional
totals_by_3rd_party_agentarray contains entries for each recognized agent app. - Each entry includes
agent_name,agent_id,user_initiated_interaction_count, andsession_count. - The
user_initiated_interaction_countfor agent apps specifically counts user-initiated job starts, distinct from the top-level field of the same name. - Metrics are aggregated by agent, collapsing multiple apps belonging to the same agent into one entry.
- This change is backward compatible; existing fields retain their shape, and the new array is omitted if no recognized agent app activity occurs.
Context
According to GitHub, agent apps have enabled teams to integrate partner agents directly into their GitHub workflows. Previously, agent activity in usage metrics was grouped, making it difficult to distinguish work done by Copilot coding agents from other agents. This update addresses the challenge of tracking multiple agents, especially as teams adopt new ones.
Why It Matters
This update provides builders and managers with granular data to understand agent adoption and usage patterns. It enables data-driven decisions regarding agent rollout and licensing, moving from assumptions to real usage metrics.
What To Do
- Visit the Copilot usage metrics API documentation to understand the new fields.
- Review the Agent apps metrics fields for full definitions.
- Note that
agent_idis the stable identifier for grouping across reporting periods. - Ensure the Copilot usage metrics policy is enabled to access these new reports.
