Disruption with some GitHub services

What to do

• Check whether your product depends on GitHub. Check your stack

  Paste a package.json, lockfile, Terraform or a list of domains at down.guide/stack to see every provider you depend on.

• Confirm your retries, queues and scheduled jobs caught up after the 3h 5m it lasted.

  Anything that failed between 2026-06-28 17:50 UTC and 2026-06-28 20:55 UTC may need a replay.

• Subscribe to GitHub's feed to hear about the next one first. RSS feed

  https://down.guide/feed?provider=github works in Slack (/feed subscribe), Discord feed bots and any RSS reader.

Updates from GitHub 5

1. investigatingJun 28, 2026, 5:50 PM UTC

   We are investigating reports of impacted performance for some GitHub services.

   provider

2. investigatingJun 28, 2026, 5:59 PM UTC

   Copilot cloud agent is experiencing intermittent problems with opening pull requests, pushing changes and replying to comments. We have identified the issue and are validating a fix.

   provider

3. investigatingJun 28, 2026, 6:31 PM UTC

   Copilot cloud agent is experiencing intermittent problems with opening pull requests, pushing changes and replying to comments. We have validated a fix and are deploying that fix now.

   provider

4. investigatingJun 28, 2026, 8:02 PM UTC

   Copilot cloud agent had been experiencing intermittent problems with opening pull requests, pushing changes and replying to comments. A fix has been deployed and we are validating that fix.

   provider

5. resolvedJun 28, 2026, 8:55 PM UTC

   From June 26, 2026 at 23:40 UTC through June 28, 2026 at 20:55 UTC, Copilot Cloud Agent was degraded. The agent could fail when reporting progress, replying to pull request comments, or opening pull requests. For affected built-in tool calls, the average error rate was approximately 8%, with hourly error rates peaking around 26%. This was due to a regression introduced during a Copilot Cloud Agent runtime deployment that caused several built-in agent tools to become unavailable. In many cases, the affected tool calls failed silently so agent jobs appeared to succeed. This monitoring gap meant it took longer than expected to identify the failure. We mitigated the incident by reverting the runtime deployment to the previously stable version. We've added monitoring and alerting for this class of tool-availability error to reduce time-to-detection. We're also adding regression tests for these built-in agent tools, and improving the shipping safety for future runtime rollouts to avoid similar issues.

   provider

Corrections

None so far.

Something wrong on this page? Report an error or send us a note. Every fix is listed here.