Apps

GitHub Copilot Logs

GitHub Copilot writes rolling diagnostic and telemetry logs under ~/.copilot/logs to record extension startup, authentication handshakes, proxy and network errors, model request timing, chat and inline completion service responses, and local agent events for troubleshooting. These plain-text log files accumulate across IDE launches and extension updates, especially when verbose logging is enabled or the extension is repeatedly retrying failed requests. Kudu removes only GitHub Copilot log files older than one week, leaving chat sessions, edit sessions, credentials, extension settings, and IDE workspace data untouched.

Why it matters

Why clean GitHub Copilot Logs?

  • Repeated auth or network retry loops can generate large daily log files, and the symptom users notice is GitHub Copilot consuming unexpected disk space in their profile folder
  • Verbose diagnostic logging during extension troubleshooting leaves behind many historical session logs, making the Copilot logs directory grow long after the original problem is gone
  • Thousands of old startup and request logs make it harder to find current errors, so users opening the log folder or attaching logs to a support case end up sifting through stale failures
  • After extension updates, old logs preserve errors from previous builds and endpoints, which can mislead troubleshooting when current Copilot behavior is actually healthy
  • Long-lived log directories create extra file indexing and antivirus scanning overhead on Windows, showing up as brief background disk activity around IDE launch or profile scans
  • Persistent transport and proxy error logs from past outages remain on disk even after service recovery, so users may keep seeing a large logs folder despite Copilot working normally again
What gets cleaned

Cache paths Kudu targets

Windows

~/.copilot/logs
Frequently asked

Common questions about GitHub Copilot Logs

Free & open source

Download Kudu and reclaim your disk space.

Available on Windows, macOS, and Linux. No account required, no feature gates, no telemetry without consent. All cleaning targets are open source and community-auditable.