Text Counter for Raycast

A Raycast extension that instantly counts words, characters, lines, sentences, paragraphs, reading time, and LLM tokens from your selected text or clipboard — with proper support for Chinese, Japanese, and Korean.

Features

• Words: space-delimited words plus CJK characters (the same convention as Word/Pages), so mixed Chinese/English text is counted correctly
• Characters: grapheme-accurate (an emoji like 👨‍👩‍👧 counts as 1), with and without spaces
• Lines / Sentences / Paragraphs: sentence detection understands both .!? and 。！？…
• Reading Time: ~200 words/min for space-delimited scripts, ~300 chars/min for CJK
• Token Counts:
  • o200k_base (GPT-4o, o-series) — exact
  • cl100k_base (GPT-4, GPT-3.5-turbo) — exact
  • Claude — estimate (Claude 3+ tokenizers are not public)
• Context Window Usage: see usage for named GPT-4o and Claude Sonnet models; the generic cl100k_base encoding has no single context window
• Cost Estimates: GPT-4o and the latest Claude Sonnet input cost, using live pricing from models.dev (cached for 24 hours; hidden gracefully when offline)

Usage

1. Select text in any app (or copy it to the clipboard)
2. Open Raycast and run "Count Tokens / Words / Lines"
3. View the statistics instantly — the top row shows a preview of the analyzed text and its source
4. Press ⏎ on any statistic to copy it, or ⌘R to refresh

Actions

• Copy Individual Counts: press ⏎ on any statistic
• Copy All Statistics: as plain text, Markdown table, or JSON
• Refresh (⌘R): re-read from selection/clipboard

Preferences

• Prefer selected text over clipboard (on by default)
• Show/hide token counts, reading time, and cost estimates

Privacy

All counting happens locally — your text never leaves your device. The only network request is an optional daily fetch of public model pricing from models.dev (no text or telemetry is sent).

License

MIT

Author

Fernando Jacob (@MatrixA) — github.com/MatrixA/counter-raycast