Disruption with some GitHub services
This summary is created by Generative AI and may differ from the actual content.
Overview
From June 26, 2026 23:40 UTC to June 28, 2026 20:55 UTC the GitHub Copilot Cloud Agent experienced a degradation. Built‑in tool calls (e.g., opening pull requests, replying to comments, reporting progress) saw an average error rate of ~8 % with peaks around 26 %. The degradation was traced to a regression introduced in a Copilot Cloud Agent runtime deployment, which rendered several built‑in tools unavailable. The issue was eventually mitigated by rolling back to the prior stable runtime version. Subsequent actions included adding monitoring and alerting for tool‑availability errors, implementing regression tests for these tools, and improving rollout safety for future deployments.
Impact
The incident caused intermittent failures for users of Copilot‑related GitHub services. Approximately 8 % of built‑in tool calls failed on average, with error spikes up to 26 %, leading to failed pull‑request operations, comment replies, and progress reporting. No explicit revenue or user‑impact metrics were provided, but the degraded functionality affected a noticeable portion of Copilot usage during the incident window.
Trigger
A regression introduced during a Copilot Cloud Agent runtime deployment caused several built‑in agent tools to become unavailable, leading to the observed error rates.
Detection
The degradation was initially identified through user reports of impacted performance and intermittent problems with pull‑request operations. A monitoring gap meant that the failure was not automatically detected, delaying awareness of the issue.
Resolution
The incident was resolved by reverting the Copilot Cloud Agent runtime deployment to the previously stable version, which restored tool availability. Additional mitigations included adding dedicated monitoring and alerting for tool‑availability errors, introducing regression tests for built‑in agent tools, and enhancing the safety of future runtime rollouts.
Root Cause
A regression in the Copilot Cloud Agent runtime deployment introduced a bug that disabled several built‑in agent tools, causing the observed degradation.
