Add disableInputSuggestions setting to project/global settings
Summary
Add a disableInputSuggestions option to .claude/settings.json (project-level) and global settings that suppresses the predictive/autocomplete suggestions shown in the chat input field.
Motivation
Input suggestions are context-aware and often anticipate the next message based on the conversation. In educational and assessment contexts, this is a significant problem: if a student is mid-lesson and the input field suggests an answer to the question just asked, it bypasses the learning entirely.
Concrete use case: I am building a Claude Code-based tutoring system. The tutor asks the student questions to assess understanding. Input suggestions frequently surface the correct answer in the input field before the student has a chance to think. There is no way to disable this at the project level today.
This is also relevant for any assessment, quiz, or interview-prep use case where the user should be producing answers independently.
Proposed setting
In .claude/settings.json:
{
"disableInputSuggestions": true
}
Project-level scope is the most important — it lets the repo owner enforce this for anyone using the project, without requiring each user to change their global settings.
Global settings support would also be useful for users who prefer a distraction-free input experience generally.
Alternatives considered
- Instructing users to dismiss suggestions manually (Escape key) — unreliable, especially for younger or less technical users
- No known workaround that enforces this at the project level
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗