[BUG] Claude Code – Agent behavior / session management

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 frequently interrupts active work with a suggestion to stop and continue in a later session, even when the task is not complete. This is a regression from prior behavior where the agent would continue working until the task was finished.

Previously, I could give Claude Code a task and it would work continuously until completion.

Now, during normal usage, the agent:

Works for ~10–20 minutes

Then interrupts with something like:

“We’ve got a lot done in this session, should we commit now and pick this up later?”

This happens before the task is complete, and without me asking to stop.

What Should Happen?

Expected Behavior
Agent continues working until:
Task is complete, or
User explicitly asks to stop
No proactive suggestion to end session mid-task
Actual Behavior
Agent interrupts workflow mid-task
Suggests committing and continuing later
Requires manual prompting to continue
Breaks flow and reduces productivity
Frequency
Happens consistently during longer tasks
Not a one-off occurrence
Impact
Interrupts development flow
Requires babysitting instead of autonomous execution
Regresses from prior “complete the task” behavior
Additional Notes
This feels like a time/session heuristic rather than task-aware behavior
The agent appears to prioritize session boundaries over task completion
This behavior did not occur previously under similar usage
Request
Remove or reduce proactive “end session” suggestions during active work
Make continuation decisions task-aware (only suggest stopping when a natural breakpoint is reached)
Allow users to opt out of session interruption prompts

Error Messages/Logs

Steps to Reproduce

Start a Claude Code session
Provide a multi-step or moderately complex task
Allow the agent to work continuously
After ~10–20 minutes, observe interruption suggesting to end session

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.86

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

I am explicitly not asking it to stop—it is choosing to stop on its own.