Pasting multibyte (Japanese/CJK) characters in terminal input causes corruption and truncation

Summary

When pasting long text containing multibyte characters (e.g. Japanese/CJK) into the Claude Code terminal input, the pasted content is corrupted or truncated. Characters are replaced with ? or similar replacement characters, and the text is often cut off mid-paste.

Environment

• Claude Code version: 2.1.87
• OS: macOS (Darwin 25.3.0, arm64)
• Terminal: Cursor integrated terminal (xterm-256color)
• Shell: zsh

Steps to Reproduce

1. Open Claude Code in a terminal
2. Copy a long string containing Japanese characters (e.g. 日本語のテキストを含む長い文章)
3. Paste into the Claude Code input prompt

Expected Behavior

The full text is received correctly, with all multibyte characters preserved.

Actual Behavior

• Some characters are replaced with ? (replacement character)
• Text is truncated mid-paste
• The input sometimes freezes and pressing Escape partially exits the frozen state, leaving only part of the pasted text

Root Cause (suspected)

The terminal input processes pasted text character-by-character rather than as a buffered block. For multibyte UTF-8 characters (Japanese = 3 bytes per character), byte sequences get split during processing, resulting in corrupted characters.

Bracketed paste mode (\e[?2004h) would solve this by wrapping paste content in escape sequences, signaling to the input handler to treat the entire paste as literal text rather than individual keystrokes.

Workaround

Currently the only reliable workaround is to use the VS Code extension's GUI input field instead of the terminal.

Request

Please enable bracketed paste mode in the terminal input handler so that:

1. Multibyte characters (Japanese, Chinese, Korean, emoji, etc.) are preserved when pasting
2. Long pastes are not truncated
3. Escape sequences within pasted text are not misinterpreted as control characters