Artificial Intelligence

GitHub Fixes Copilot Metrics Bug Affecting Agent Activity

A recent change moving Copilot agent sessions to the Copilot SDK caused usage metrics to undercount agent activity while overall Copilot usage continued to grow.

Radwan AbdulhadiEditor and technology writer

Published: October 7, 2026 · Updated: October 7, 2026 · 4 min read

GitHub Fixes Copilot Metrics Bug Affecting Agent Activity — Image: GitHub Changelog

The Cause Behind Falling Agent Metrics

Developers monitoring their GitHub Copilot usage metrics might have noticed an unusual discrepancy: overall Copilot usage continued to grow, but agent activity and agent lines of code appeared to be falling. According to the official [GitHub Changelog](https://github.blog/changelog/2026-10-06-update-your-ide-to-restore-agent-activity-in-copilot-usage-metrics/), this issue was triggered when several IDEs recently moved Copilot agent sessions to the Copilot SDK.

Because these sessions failed to identify which integrated development environment they originated from, the usage metrics system could not attribute them correctly. As a result, the majority of this activity was omitted from reports, while some was incorrectly counted as Copilot CLI activity. Only IDE versions utilizing the Copilot SDK for agent mode are impacted by this attribution bug, meaning developers remaining on earlier versions are still being counted normally.

Rollout Schedule and Developer Actions

The fix for this reporting error is already available in Visual Studio Code. Other compatible platforms and extensions will receive the fix in upcoming releases, with the complete rollout expected to finish by November 2026. Developers who update their software will immediately see their agent activity tracked accurately once again in the Copilot usage metrics dashboard and API.

Organizations and engineering teams managing their IDE versions centrally can plan their deployment strategies to move developers directly to the patched versions. For environments utilizing [supported IDEs](https://docs.github.com/copilot/concepts/billing-and-usage/copilot-usage-metrics/copilot-metrics#supported-ides), updating client applications as soon as the fixed versions become available is crucial because past data cannot be recovered.

Impact on Billing and Historical Data

GitHub confirmed that billing remains completely unaffected by this telemetry oversight. The problem strictly involved how agent activity was attributed within usage metrics reports rather than what customers were charged for their usage.

However, the data gap persists for any developer operating on an affected IDE version until they complete their update. Until then, agent interactions and agent lines of code—such as loc_added_sum and loc_deleted_sum for agent_edit—will remain undercounted across enterprise, organization, and user reports for both 1-day and 28-day views. GitHub has stated that missing data cannot be backfilled because activity from affected versions lacks IDE identification, meaning metrics will experience a gradual recovery rather than a single sudden jump as users update.

Inflated Copilot CLI Metrics

Another side effect of the attribution bug involved Copilot CLI metrics appearing inflated, as activity originating from other SDK-based clients was mistakenly grouped into CLI reporting. This distortion will clear up organically as developers update their client software.

Despite this inflation, earlier Copilot CLI metrics cannot be retroactively corrected because that misattributed activity cannot be cleanly separated from genuine Copilot CLI use. Notably, regular users of the Copilot CLI do not need to perform any updates to resolve this.

Understanding Client-Side Telemetry Gaps

Detailed usage metrics such as feature breakdowns, language stats, model selections, and lines of code depend heavily on the telemetry sent directly by each IDE. Although GitHub records separate server-side data when handling requests to show who was active, that server data cannot observe editor-side actions. Disagreements between reports and server data typically point back to client-side conditions.

Common reasons for discrepancies include disabled telemetry settings within the editor, network proxies or firewalls blocking the telemetry endpoint, outdated extensions, or changes in how telemetry transmission occurs. GitHub continues to refine its approach, using server-side data to [count active users that client telemetry misses](https://github.blog/changelog/2026-06-15-copilot-usage-metrics-now-include-more-of-your-active-users/) and to [identify the IDE for those users](https://github.blog/changelog/2026-07-02-improved-accuracy-and-coverage-in-copilot-usage-metrics-reports/).

Because certain metrics like accepted suggestions and lines of code require direct editor feedback, client-side gaps will not disappear entirely. To maintain comprehensive reporting, engineering managers are advised to enforce minimum versions using device management tools, keep IDE telemetry enabled, and review per-user totals to spot outdated clients.