Copilot使用状況メトリクスAPIがエージェントアプリのアクティビティに対応Copilot usage metrics API adds agent app activity

GitHub has extended its Copilot usage metrics API so that it now reports activity generated by agent apps, giving organizations a unified view of how AI agents from partners such as Claude and Codex are being used across their GitHub workflows. The update matters because agent-based tools have spread quickly through developer pipelines, and until now the data describing their use largely lived outside the reporting surface that many teams already rely on to track Copilot adoption.

Since agent apps arrived on GitHub, teams have been able to run agents from partners like Claude and Codex directly in their GitHub workflows, delegating tasks that go beyond inline code completion. These agents can operate on issues, pull requests, and other parts of a repository, acting more autonomously than the suggestion-based assistance that defined earlier versions of Copilot. As adoption grew, administrators had limited standardized ways to quantify that activity alongside their existing Copilot metrics.

The change addresses that gap by folding agent app activity into the same API that already aggregates Copilot usage data. In practice, this means the numbers describing how agents are invoked and used can now be pulled from a single, consistent endpoint rather than assembled from separate sources. For organizations managing several AI tools at once, that consolidation is likely to simplify reporting and make it easier to compare usage patterns across different assistants and agents.

The Copilot usage metrics API is a REST interface that returns aggregated, organization- and enterprise-level data about how Copilot features are being used. It has historically exposed measures such as active and engaged users, along with breakdowns that help administrators understand where and how the tool is adopted. Extending it to cover agent apps continues a pattern of broadening the API's scope as GitHub's AI product surface expands beyond code completion into chat, code review, and now agent-driven workflows.

The distinction between partner agents is worth noting for context. Claude is developed by Anthropic and Codex is associated with OpenAI, and both are examples of third-party agents that GitHub has made available to run within its platform. By treating their activity as something to be measured through the Copilot metrics API, GitHub appears to be positioning the API as a general-purpose reporting layer for agentic work in its ecosystem, not only for its own first-party features.

For teams responsible for governance, the update carries practical weight. Metrics APIs of this kind are commonly used to build adoption dashboards, justify licensing spend, identify teams that may need enablement, and monitor whether AI investments are translating into measurable engagement. Adding agent activity gives those workflows a more complete picture, since a growing share of AI-assisted work is expected to shift toward agents that can carry out multi-step tasks rather than simply suggesting the next line of code.

It is useful to keep in mind how these metrics typically behave. Data returned by the Copilot metrics API is generally aggregated rather than tied to individual developers, and access is usually gated behind organization or enterprise administrative permissions. Reporting endpoints of this type often apply minimum thresholds before returning results, and figures can lag real-time activity because they are compiled on a periodic basis. Teams integrating the new agent data should confirm the exact fields, granularity, and refresh cadence in GitHub's documentation before building on top of it.

The announcement fits into a broader industry move toward measuring, not just deploying, AI coding tools. As vendors including GitHub, Anthropic, and OpenAI push agents deeper into everyday engineering work, buyers are increasingly asking for evidence of value and clearer controls. Instrumentation like this, surfacing which agents are being used and how often, is a prerequisite for that kind of oversight. For organizations already standardized on GitHub, having agent activity reported through the same API they use for Copilot is likely to lower the barrier to tracking these tools consistently as their agent usage continues to grow.