### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement # Feature Request: Predictive Text / Smart Compose for User Input ## Summary Add real-time text prediction and autocomplete for natural language input in Claude Code's interactive mode, similar to WhatsApp, Gmail's Sm

[FEATURE] Predictive Text / Smart Compose for User Input

Resolved 💬 2 comments Opened Feb 15, 2026 by apraka16 Closed Mar 15, 2026

Preflight Checklist

[x] I have searched existing requests and this feature hasn't been requested yet
[x] This is a single feature request (not multiple features)

Problem Statement

Feature Request: Predictive Text / Smart Compose for User Input

Summary

Add real-time text prediction and autocomplete for natural language input in Claude Code's interactive mode, similar to WhatsApp, Gmail's Smart Compose, or mobile keyboard predictive text.

Current Behavior

Claude Code currently provides:

Prompt suggestions after Claude responds (static follow-up suggestions)
File path autocompletion with @ prefix
Bash command history autocompletion with ! prefix

However, there is no real-time text prediction while the user is actively typing their message.

Requested Feature

Implement intelligent text prediction that suggests the next word or completes sentences as users type their messages to Claude. This would work similarly to:

WhatsApp - Word suggestions appear above the keyboard as you type
Gmail Smart Compose - Gray text appears inline showing predicted sentence completion
Mobile keyboards - Next-word predictions in real-time
GitHub Copilot - Inline suggestions in code editors

Use Case & Benefits

1. Improved Typing Speed

Users frequently interact with Claude using natural language. Text prediction would significantly speed up message composition, especially for:

Longer, detailed prompts
Repetitive phrases (e.g., "Can you help me...", "Please update the...", "Let me know if...")
Common development terminology

2. Reduced Typing Friction

Many users are accustomed to predictive text from messaging apps and email clients. Its absence in Claude Code creates friction when transitioning from those apps.

3. Better Mobile/Touch Experience

For users accessing Claude Code via terminals on tablets or touch devices, predictive text is essential for efficient input.

4. Accessibility

Predictive text helps users with:

Mobility impairments who find typing challenging
Dyslexia or other conditions where word completion assists communication
Non-native English speakers who benefit from word suggestions

Proposed Implementation

Option 1: Inline Suggestions (Preferred)

Display predicted text inline as grayed-out text after the cursor:

User types: "Can you help me refact"
Display:     "Can you help me refactor this component"
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
             (grayed out, press Tab to accept, keep typing to ignore)

Option 2: Suggestion Bar

Show 2-3 word suggestions above the input area (similar to mobile keyboards):

Suggestions: [refactor] [review] [update]
User input:  Can you help me ref_

Option 3: Hybrid Approach

Combine both methods - show next-word options in a bar, with full sentence completion inline.

Suggested Keybindings

Tab - Accept current suggestion
Right Arrow - Accept one word from suggestion
Escape - Dismiss suggestion
Continue typing - Ignore suggestion and keep composing

Configuration Options

{
  "textPrediction": {
    "enabled": true,
    "style": "inline" | "bar" | "hybrid",
    "model": "local" | "api",
    "minCharacters": 3,
    "maxSuggestions": 3
  }
}

Technical Considerations

Privacy & Performance

Local-first approach: Consider using a lightweight local language model for predictions to avoid sending user input to external services
Opt-in: Make this feature optional via settings
No logging: Ensure predictions don't send data externally unless explicitly configured

Model Options

Local n-gram or neural model (fast, private, offline)
Claude API integration (more intelligent but requires API calls)
Hybrid: Local for common words, API for complex technical completions

Similar Implementations

GitHub Copilot CLI - Provides command suggestions
Fig (now Amazon Q) - Autocomplete for CLI commands
Warp Terminal - AI-powered command suggestions
Google Gmail - Smart Compose for email

Impact

This feature would significantly improve the user experience for Claude Code's interactive mode, making it feel as modern and responsive as other text-based interfaces users interact with daily.

Questions for Discussion

Should predictions be context-aware based on previous conversation history?
Should the system learn from user's typing patterns over time?
What should be the default behavior (on/off) for new installations?
Should there be different prediction styles for code vs. natural language?

---

Would love to hear thoughts from the team on this! Happy to provide more details or help with testing if this gets prioritized.

Proposed Solution

Proposed Implementation

Option 1: Inline Suggestions (Preferred)

Display predicted text inline as grayed-out text after the cursor:

User types: "Can you help me refact"
Display:     "Can you help me refactor this component"
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
             (grayed out, press Tab to accept, keep typing to ignore)

Option 2: Suggestion Bar

Show 2-3 word suggestions above the input area (similar to mobile keyboards):

Suggestions: [refactor] [review] [update]
User input:  Can you help me ref_

Option 3: Hybrid Approach

Combine both methods - show next-word options in a bar, with full sentence completion inline.

Alternative Solutions

_No response_

Priority

High - Significant impact on productivity

Feature Category

Interactive mode (TUI)

Use Case Example