[BUG] IME (ATOK) candidate window overlaps input area on Windows Terminal

Status Open
Reported on v2.1.191
Maintainer reply None cached
Activity 4 comments · opened Jun 25, 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 typing Japanese text using ATOK IME in Claude Code's input prompt,
the IME candidate/conversion window overlaps with the input area,
making it difficult to see what is being typed.

What Should Happen?

The IME candidate window should appear directly above or below the
input cursor without overlapping the input text, as it does in
normal terminal input (outside of Claude Code's TUI).

Error Messages/Logs

Steps to Reproduce

  1. Launch Claude Code in Windows Terminal on Windows 11
  2. Switch to ATOK (Japanese IME)
  3. Type Japanese text requiring conversion (e.g. "かんきょう" → "環境")
  4. Observe the position of the IME candidate window

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.191 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Windows Terminal

Additional Information

  • Windows 11 26H2 (Build 26300)
  • Windows Terminal 1.24.11321.0
  • IME: ATOK (JustSystems)
  • The same issue likely occurs with MS-IME and Google Japanese Input,

as the root cause appears to be Claude Code's TUI (React Ink) hiding
the real terminal cursor position from the OS IME subsystem.

  • Related issues: #1547, #25186, #19207, #35307

<img width="1423" height="354" alt="Image" src="https://github.com/user-attachments/assets/cba0f327-a656-4b2e-a032-eb6ba9e60732" />

View original on GitHub ↗

4 Comments

github-actions[bot] · 2 months ago

Found 3 possible duplicate issues:

  1. https://github.com/anthropics/claude-code/issues/63091
  2. https://github.com/anthropics/claude-code/issues/61385
  3. https://github.com/anthropics/claude-code/issues/35307

This issue will be automatically closed as a duplicate in 3 days.

  • If your issue is a duplicate, please close it and 👍 the existing issue instead
  • To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

tricksterzero · 2 months ago

Not a duplicate. All three referenced issues are closed:

  • #63091 — auto-closed as a duplicate of #61385
  • #61385 — closed as fixed by a collaborator (bogini, 2026-05-27)
  • #35307 — auto-closed as stale/inactive, then locked

If the fix merged for #61385 fully resolved this class of bug, I should not be seeing it on v2.1.191. This is either a regression or an
incomplete fix.

Additionally, #61385 was specifically about the Agent View → session transition path. My report is about normal interactive input — no
Agent View involved.

Since all three issues are closed (and #35307 is locked), there is no existing open issue to upvote. This issue serves as a fresh repro on
the latest version.

tricksterzero · 1 month ago

Still reproducing on Claude Code 2.1.220 (Windows Terminal, ATOK) -- same overlap between the IME candidate window and the input area as originally described. Flagging this to prevent stale auto-close; the fix in #61385 was specific to the Agent View to session transition path and doesn't cover this normal interactive-input case.

tricksterzero · 10 days ago

Additional data point on the root cause: on Windows Terminal (Claude Code 2.1.238, ATOK), the candidate window's vertical position tracks the caret correctly (confirmed across different rows). However, the horizontal position is stuck at the terminal's left edge (column 0) regardless of caret column — reproduced by typing a long line of digits and triggering the emoji/candidate palette mid-line; the popup rendered flush-left while vertically aligned with the input row.

This is narrower than the "real cursor fully hidden, falls back to (0,0)" explanation in the issue body — Y is not falling back, only X is. Might point at a column-calculation bug specific to Windows/ConPTY (e.g. wide-character width miscount in the column offset, or the column component of the cursor-position escape sequence not being emitted/interpreted correctly) rather than the cursor being hidden entirely.