[BUG] [BUG] Claude downtime mid-session caused all uncommitted work to be lost

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Claude Code experienced downtime in the middle of an active coding session. I was in the process of implementing significant features (landing page redesign, template selector UI overhaul) across multiple files. Because Claude went down before I could commit the changes, everything was lost — the modified files reverted and there is no recovery path.

What Should Happen?

Claude Code should either:

1. Auto-save a patch/diff of all file changes made during a session so work can be recovered after downtime
2. Checkpoint commits periodically during long sessions
3. At minimum, on reconnect, show the user what was changed and offer to re-apply or commit it

Error Messages/Logs

Steps to Reproduce

1. Start a long coding session with Claude Code (Accept Edits ON)
2. Claude makes changes across many files
3. Claude goes down mid-session before changes are committed
4. Reconnect — all file changes are gone, git history is unchanged

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

unknown (downtime happened mid-session, not a version regression)

Claude Code Version

 2.1.109 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Critical impact. I lost several hours of work and millions of tokens. My token budget is now nearly depleted and I have to rebuild everything from scratch. There is no way to recover what was built. I need refund of tokens and limit renewal/

This isn't a model behavior issue — it's an infrastructure/reliability issue. When a paid service goes down mid-session, users should not lose their work. At minimum there should be a session recovery mechanism or local change log.