[BUG] Conversation .jsonl not written to disk on Windows 11 — v2.1.70 regression

Status Fixed / completed

Maintainer reply None cached

Activity 4 comments · opened Mar 6, 2026 · closed Mar 7, 2026

Describe the bug

After updating to v2.1.70 (released 2026-03-06), conversation data is no longer persisted to disk. Session directories are created under ~/.claude/projects/<project>/<session-id>/ but contain no main conversation .jsonl file — only tool-results/ and subagents/ subdirectories.

Both --resume <session-id> and --continue fail:

$ claude --continue
No conversation found to continue

$ claude --resume 8c47cb6e-2586-485c-b4c3-449ade62e27a
No conversation found with session ID: 8c47cb6e-2586-485c-b4c3-449ade62e27a

This is a regression. I am a heavy Claude Code user and --resume / --continue have been working fine until today's update. This is not the long-standing issue from v2.1.31 (#26123) — that bug never affected me.

Evidence

Session directory exists but has no conversation transcript:

$ ls -la ~/.claude/projects/D--python-optiflightpoc/8c47cb6e-2586-485c-b4c3-449ade62e27a/
total 12
drwxr-xr-x 1 mike 197121 0 Mar  6 21:07 .
drwxr-xr-x 1 mike 197121 0 Mar  6 21:07 ..
drwxr-xr-x 1 mike 197121 0 Mar  6 21:07 tool-results

No .session_index file exists either. Only subagent .jsonl files are present across all session directories — zero main conversation .jsonl files.

Platform

OS: Windows 11 (native, not WSL)
Shell: MSYS2 / Git Bash
Claude Code: v2.1.70
Previous version: was working correctly

Steps to reproduce

Update to v2.1.70 on Windows 11
Start a new claude session, have a conversation with tool use
Exit the session
Run claude --continue → "No conversation found to continue"
Run claude --resume <session-id> → "No conversation found with session ID: ..."
Inspect ~/.claude/projects/<project>/<session-id>/ — directory exists but has no .jsonl

Expected behavior

Main conversation .jsonl should be written to the session directory, and --resume / --continue should work as they did prior to v2.1.70.

View original on GitHub ↗

4 Comments

github-actions[bot] · 7 months ago

Found 3 possible duplicate issues:

This issue will be automatically closed as a duplicate in 3 days.

If your issue is a duplicate, please close it and 👍 the existing issue instead
To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

geojakes · 7 months ago

I don't think it's just a problem with 2.1.70, I went back to 2.1.44 and it was still there. Essentially, the short task (requires 1 step for the LLM) end up running so fast that a sigterm is sent out even before the session gets written into persistence. See details here -> https://github.com/anthropics/claude-code/issues/31613

MikeWise2718 · 7 months ago

Ok, in fact I had recently cleanupPeriodDays to 0. Using a large number instead (I went with 730), caused conversations to be stored again. So this was in fact issue 23710.

Pretty strange bug on the face of it.

I seem to have lost a lot of saved conversations though. Not good.