G: TL;DR:
Yes — what you’re asking for does exist, but it’s rare because Japanese typing is fundamentally built around IMEs. Your best matches are standalone dictionary/IME frontends or self-contained apps that embed an IME without system-wide takeover. The key names to look at are Mozc (standalone-ish usage), uim-anthy, and actual apps like Gjiten/Kiten used as “typing sandboxes.”

---

What you’re trying to avoid (and why it’s hard)

Japanese typing ≠ normal keyboard input. It requires:

• phonetic input (romaji/kana)
• conversion engine (kana → kanji suggestions)

That conversion layer is what IMEs do. On Linux, things like entity["software","Fcitx","Linux input method framework"] or IBus are just frameworks that inject into the whole system, which is exactly what you don’t want. citeturn0search11

So your constraint is basically:

“IME functionality, but contained inside one program.”

That’s not the default design, so you have to cheat a bit.

---

Option 1 — “Contained IME” (best practical compromise)

Use Mozc, but only inside a single app context.

• entity["software","Mozc","Google open-source Japanese IME"] is the Linux version of Google Japanese Input citeturn0search12
• You can run it with:
  • minimal integration (no global hotkeys)
  • tied to a single input method launcher (like fcitx5-mozc, but only active when you launch it manually)

Why this fits your goal:

• You don’t have to switch your whole OS to Japanese input
• You can treat it like a “mode” you open when needed
• Close it → system goes back to normal

Reality check:
Still technically an IME, but you can sandbox its use enough that it behaves like a tool instead of a system mutation.

---

Option 2 — Dictionary apps as “typing environments”

These are not IMEs, but they give you a contained textbox where Japanese input works.

entity["software","Gjiten","Linux Japanese dictionary app"]

• Standalone GUI app
• Lets you type kana/kanji and search
• Has kanji lookup + handwriting support citeturn0search10

entity["software","Kiten","KDE Japanese study tool"]

• Similar idea, KDE-based
• Input box + kanji tools + vocab lookup citeturn0search13

How people actually use these:

• Open the app
• Type Japanese inside it (using a minimal IME or paste workflow)
• Treat it like a scratchpad

Why this works for you:

• Completely self-contained UI
• No system-wide weirdness
• Feels like “a Japanese typing app”

Downside:

• Still needs some input method behind the scenes for kana/kanji conversion

---

Option 3 — Terminal-based isolation (cleanest separation)

This is the closest to what you actually want.

Use:

• a terminal IME (like uim or fcitx in terminal-only mode)
• inside a single terminal window

Then:

• run something like nano, vim, or even a simple text UI
• only that terminal session has Japanese input

Effect:

• The rest of your desktop is untouched
• You’ve effectively sandboxed the IME manually

---

Option 4 — No IME at all (pure self-contained, but limited)

If you truly want zero system integration, you’re stuck with:

• romaji → kana converters (web tools or scripts)
• copy/paste workflow
• or custom scripts (e.g., Python + MeCab wrappers)

This avoids IMEs entirely, but:

• no live kanji conversion
• clunky for real typing

---

Bottom line (no fluff)

• What you want isn’t the normal Linux model, but it’s doable.
• The cleanest setups are:
  1. Mozc used only when launched (semi-contained)
  2. Terminal sandbox IME (most controlled)
  3. Dictionary apps as typing shells (most “app-like”)

If you’re trying to avoid system weirdness, Option 3 (terminal isolation) is the least invasive and most predictable.