中文

Prompt Diff & Version Compare

Compare prompt revisions by line, word, or grapheme with real token deltas and optional local versions.

Local prompt version laboratory

Prompt diff & version workspace

No uploads · No default body persistence
160 UTF-16 charactersEdits stay in memory unless you explicitly save
or drop a local file here (up to 1 MB)
182 UTF-16 characters
or drop a local file here (up to 1 MB)
Original characters160
New characters / delta182 +22
Original tokens—
New tokens / delta—
Counting locally…
0 added · 0 removed · 0 changed groups
Inspect every segment

Comparison

Both versions are empty. Enter text or load the example to begin.
Explicit local saves

Versions

0 / 40

Only “Save version” writes both prompt bodies to this browser’s localStorage. Unsaved drafts remain in this page’s memory only.

Delete, clear, and undo never clear the editors.
Explicit sharing

Generate a link on demand

No prompt enters a URL before you click. Fragments are not encrypted; they are normally omitted from HTTP requests but remain in links you share, browser history, or screenshots. Up to 12,000 source characters and 16,000 encoded characters; export longer content instead.

Diffing, tokenization, import, and export run locally; no model or network API is called.
About this tool

Made for the task at hand.

Review prompt changes without losing the text you are still editing. Compare verbatim original and revised inputs at line, word, or grapheme level, measure exact counts for a chosen tiktoken encoding, and opt into named browser-local snapshots only when needed.

How to use

  1. Paste or import original and new prompts, then choose line, word, or character comparison and any comparison-only ignore options.
  2. Review accessible highlighted changes and character/token deltas using the selected real encoding; export the comparison or current version.
  3. Explicitly save, rename, duplicate, switch, delete, or undo local versions. Generate a size-limited share fragment only when you choose, or use files for long prompts.
Your work stays yours.
Tool inputs are processed locally in your browser and are never sent to KitVerse servers. Clear the workspace when using a shared device.
Common questions
Are prompt drafts saved automatically?

No. Edits stay in page memory and survive version switching, but not a refresh. Only Save version explicitly writes both prompt bodies to localStorage. Delete, undo, and Clear saved versions do not clear the editors.

Are token counts exact?

Counts use the real o200k_base or cl100k_base encoder you select. They are exact for models using that encoding and estimates otherwise; chat wrappers and provider overhead are excluded.

How do share links handle privacy and size?

A #share= fragment is generated only after you click Share. It is encoded, sometimes compressed, but never encrypted. Fragments are normally omitted from HTTP requests, yet remain visible in the link and browser history. Links are limited to 12,000 source characters and 16,000 encoded characters; longer prompts require file export.

What are the processing limits?

Each editor accepts up to 500,000 UTF-16 characters and imports up to 1 MB. Up to 40 named versions can be stored subject to browser quota. Long diffs run in a Worker and the visual diff is capped at 4,000 segments; complete source text remains available for export.

Related tools · AI & LLM

New

Tokenizer Playground

See exactly how text becomes tokens, then compare the cost across popular LLMs.

Free

Token Counter

Count characters and LLM tokens with a lightweight, private text workspace.

Free

Token Cost Calculator

Estimate input and output spend with transparent per-million-token pricing.