Textkiln is a collection of writing and text tools built on three convictions.
1. Your text is yours
Writers paste drafts, contracts, student essays, medical letters and client work into online tools every day, usually without knowing where it goes. Every Textkiln tool runs in your browser. Text is not uploaded, logged or used to train anything. Drafts are kept only in your browser’s local storage so they survive a refresh, and you can clear them at any time. Many tools keep working offline once loaded.
2. Precise beats approximate
A good tool is one you can trust without checking. That means following the actual standards: Unicode text segmentation for words and characters, each social platform’s real counting rules, pixel metrics for search snippets, the published readability formulas, and the style guides’ own capitalisation rules. Where a result is an estimate (page counts, syllables, pattern-based style checks) the tool says so.
3. No rented intelligence
Textkiln does not depend on hosted AI services. The grammar checker is Harper, an open-source engine compiled to WebAssembly; language detection uses franc; Markdown, diffs and PDFs use mature open-source libraries; the style checks use transparent rules and curated word lists we maintain ourselves. That keeps the tools fast, free to use, private, and impossible for a third party to switch off.
How we build and check the tools
- Standards first. Counting follows Unicode Standard Annex #29 (text segmentation); social counters follow each platform’s published rules (for X, its open-source twitter-text configuration); SMS follows the GSM 03.38 alphabet; citations follow the current APA, MLA and Chicago manuals; readability uses the formulas as originally published.
- Tested against independent answers. Every engine has automated tests whose expected values were worked out by hand or taken from the standard (not copied from the tool’s own output) and every tool is exercised in a real browser before release.
- Sources on the page. Each tool explains its method and cites research where numbers come from research (for example, Brysbaert’s 2019 meta-analysis for reading speed).
- Honest limits. Where a result is an estimate (syllables, page counts, pattern-based style checks) the tool says so.
- Reviewed regularly. Platform limits, style-guide editions and Unicode data are reviewed at least twice a year; each tool page shows when it was last reviewed.
How Textkiln is funded
Textkiln is free to use without an account. We plan to support it with a small number of clearly labelled, non-intrusive advertisements placed away from your writing, and optional extras for power users. We will never sell or share what you write (we never receive it in the first place).
Credits
Textkiln stands on the shoulders of open-source projects including Harper (Apache-2.0), marked, Turndown, jsdiff, PDF.js, fflate, franc, syllable, stemmer and entities, and on the Unicode Character Database. Typefaces: Instrument Sans and JetBrains Mono, both under the SIL Open Font License.
Contact
Found a bug, a wrong count or a rule we’ve misread? We want to hear about it. (Contact address to be added before launch.)