[BUG] Claude enters plan mode automatically without waiting for user approval
Status Open
Reported on v2.1.27
Maintainer reply None cached
Workaround ✓ Mentioned in thread ↓
Activity 7 comments · opened Jan 31, 2026
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?
When Claude determines it should enter plan mode, it immediately enters plan mode without waiting for user confirmation. Even when Claude explicitly asks "Would you like me to enter plan mode?", it doesn't wait for a response before switching.
What Should Happen?
Claude should prompt the user and wait for approval before entering plan mode, allowing the user to decline if they prefer Claude to proceed without planning.
Error Messages/Logs
Example 1 - Statement that should still require approval:
This is a straightforward fix. I should enter plan mode to design this properly.
● Entered plan mode
Claude is now exploring and designing an implementation approach.Example 2 - Question that was never answered:
Would you like me to enter plan mode to design a comprehensive reorganization with proper 50+ ID ranges for each category? This will involve:
1. Defining a new EventId allocation scheme
2. Updating all EventIds across the logging extension files
3. Ensuring no overlaps and consistent grouping
● Entered plan mode
Claude is now exploring and designing an implementation approach.Steps to Reproduce
- Start a new Claude Code session
- Ask Claude to do a complex task that will trigger plan mode prompt
- Observe that Claude enters plan mode without prompting for approval
Claude Model
Opus
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.27
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
PowerShell
Additional Information
_No response_
7 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
I agree. This should be addressed. I'm seeing "plan mode" being activated even for simple tasks which unnecessarily costs tokens.
+1 - Same issue here. Model calls EnterPlanMode proactively without waiting for approval, even with CLAUDE.md instructions saying not to.
I managed to disable this behaviour by running Claude Code using
code --disallowedTools "EnterPlanMode"My main issue is that I'm keeping the planning documents in the project directory tree so I can read through them (in a formatted IDE) and make changes to them before either starting them or deferring them for a later time. Planning Mode creates plans in a hidden directory (~/.claude/plans) with
stupid_silly_naming_convention.mdand no structured directory tree, making revisiting plans a pain.My workaround will allow Claude to use planning mode but output the results to the correct locations. I'm using a pre-tool hook in the main settings files:
```cmd=$(jq -r '.tool_input.command // ""')
echo "$cmd" | grep -qE '(chmod|mkdir).*\.claude/plans' && {
echo 'Blocked: Do not use .claude/plans. Write the planning document in the
documents/planning directory of the current working directory with a sensible
name following the existing naming convention.' >&2
exit 2
} || exit 0
The opposite of this is an issue too: it shows a plan but doesnt enter plan mode when asked to "make a plan"; instead, it shows some bullet points without the specificity of the actual plan mode.
This happens to me when I have relatively small tasks. I think it saves some tokens. Not sure if it is a feature or a bug.