[BUG] Claude Desktop: Chat rollbacks/disappearing responses when using MCP tools

Status Fixed / completed

Maintainer reply None cached

Activity 4 comments · opened Jan 8, 2026 · closed Jan 9, 2026

Preflight Checklist

[x] I have searched existing issues and this has not been reported for Claude Desktop specifically
[x] This is a single bug report
[x] I am using the latest version of Claude Desktop

What is Wrong?

Claude Desktop responses disappear immediately after completion when using MCP tools. The conversation "rolls back" to the state before the response, losing all generated content. This happens intermittently but frequently during longer sessions with MCP tool usage.

This issue has been reported in community discussions for 6+ months without resolution.

Log Analysis

Analysis of ~/.config/Claude/logs/claude.ai-web.log reveals the following errors correlating with rollbacks:

| Error | Count |
|-------|-------|
| Failed to invalidate conversation tree | 13 |
| Request timeout | 18 |
| Tool result is too large. Maximum size is 1MB | 5 |
| Internal server error | 4 |

The Failed to invalidate conversation tree error appears to be the direct cause - the app fails to sync conversation state after certain errors occur.

What Should Happen?

Responses should persist after completion. If an error occurs, the user should see an error message rather than silent data loss.

Steps to Reproduce

Start a Claude Desktop session with an MCP server connected
Have an extended conversation with multiple MCP tool calls
Eventually a response will complete visually but then disappear
The conversation rolls back to the previous state

More likely to occur:

After longer conversations
After large MCP tool responses
When Claude Desktop is not in focus during generation

Environment

Claude Desktop: 0.14.10
Platform: Linux (Arch-based, CachyOS)
MCP Server: Desktop Commander 0.2.28 (single MCP, installed via npx)
Subscription: Max

Suggested Fixes

Add retry logic for conversation tree sync failures
Show error messages instead of silent rollbacks
Cache responses client-side before confirming server sync
Increase/remove the 1MB tool result limit or handle gracefully

Additional Context

This is filed here because there is no public issue tracker for Claude Desktop. The issue significantly impacts productivity for paying customers who rely on MCP integrations.

View original on GitHub ↗

4 Comments

github-actions[bot] · 9 months ago

Found 2 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

AtlAntA118 · 9 months ago

Closing as duplicate of #10065. Added our Linux log analysis and findings there.

Budgulick · 9 months ago

Additional Data - 18 Incidents Over 8 Days (Windows 11)

I'm experiencing the identical issue on Windows 11 with Desktop Commander MCP. Adding my data to help establish this as a systemic problem.

My Stats:

Platform: Windows 11, Claude Desktop (latest)
- MCP: Desktop Commander only (minimal config)
- - Incidents: 18 session-destroying events from Jan 6-13, 2026
- - - Worst Day: Jan 12 - 7 resets in one day

Error Pattern (matches your findings exactly):

[COMPLETION] Request failed Error: Request timeout
[COMPLETION] message_store_sync_loss  
Failed all attempts to invalidate conversation tree
"This conversation could not be found"

New Pattern - Jan 12: "Overloaded" Errors
Starting Jan 12, I'm seeing heavy "Overloaded" errors (not previously observed):

12:08:19 [COMPLETION] Request failed Error: Overloaded
12:08:51 [COMPLETION] Request failed Error: Overloaded
12:09:31 [COMPLETION] Request failed Error: Overloaded
... (10 consecutive failures, exponential backoff)
12:21:34 message_store_sync_loss

My Log Analysis (from claude.ai-web.log):
| Error Type | Count |
|------------|-------|
| Request timeout | 12 |
| Overloaded | 3 clusters (Jan 12+) |
| Internal server error | 4 |
| Network error | 2 |
| message_store_sync_loss | 18 (all incidents) |
| Failed to invalidate conversation tree | 18 |

Support Timeline (No Resolution):

Jan 6: First incidents, thumbs-down feedback
- Jan 8: Support ticket with full logs
- - Jan 8: Received MCP misdiagnosis response
- - - Jan 8: Sent technical rebuttal proving server-side cause
- - - - Jan 11: Multiple escalation emails (5 incidents that day)
- - - - - Jan 13: 18 total incidents, still no engineering acknowledgment

Conclusion: This confirms the issue affects multiple platforms (Linux in your report, Windows in mine) with the same error fingerprint. The "Overloaded" errors on Jan 12 suggest capacity issues on Anthropic's infrastructure.

Happy to provide full logs or additional data.

github-actions[bot] · 8 months ago

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.