Add autoContinue setting to skip continuation prompts
Problem
When Claude Code hits output length limits during long responses, it pauses to inform the user "I need to continue" and waits for acknowledgment before proceeding. While this provides transparency, it interrupts workflow for tasks that involve batch processing or creating multiple files sequentially.
Use Case
I work with a large documentation project (Series Bible for a fiction series) where Claude Code frequently:
- Processes chat logs line-by-line to extract character details
- Creates/updates multiple character profiles, medical references, and relationship files
- Updates tracking documents after each batch of changes
These workflows are intentionally thorough and detailed (per project requirements), which means continuation pauses happen regularly. Each pause requires manual intervention to proceed, breaking focus and slowing down the work.
Proposed Solution
Add an autoContinue setting (similar to existing acceptEdits and bypassPermissions modes) that allows Claude Code to continue automatically when hitting output limits.
This could be implemented as:
Option 1: Session command
/autoContinue
Enables automatic continuation for the current session.
Option 2: Settings file
{
"autoContinue": true
}
Option 3: Both
Support both session command and persistent setting.
Benefits
- Maintains workflow continuity for batch operations
- Reduces manual intervention for known-safe, long-running tasks
- Consistent with existing permission modes (acceptEdits, bypassPermissions)
- Optional feature that doesn't affect users who prefer current behavior
Alternatives Considered
- Using extended context models ([1m] suffix) - doesn't eliminate continuation pauses, just delays them
- Breaking tasks into smaller chunks - defeats the purpose of thorough, detailed processing requirements
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗