6.1k stars · MIT · v3.20.0 (2026-09-02) · Track this in Scout
A linter for your voice — the difference between a site that sounds like a company and one that sounds like nine separate afternoons.
▶Repo detailsthe review · specs · pros & cons
What it is
A markup-aware command-line linter for prose. You point it at a style — Microsoft's, Google's, or a set of YAML rules you write yourself — and it flags passive voice, banned phrases, heading capitalisation and inconsistent product terms across Markdown, HTML and MDX, while ignoring the code and front-matter around them.
What it is good for. LanguageTool tells you a sentence is wrong. Vale tells you it is not yours. Writing docs, blog posts and video descriptions over months, the failure is not grammar, it is drift: "sub-topic" on Monday and "subtopic" on Friday, "map" in the hero and "graph" in the docs. Those wobbling terms are exactly the strings an AI answer engine would otherwise quote a site by, so terminology consistency is an AI-search asset rather than a style hobby — the same argument geo-optimizer-skill (Ed.8 #3) makes about structure, applied to vocabulary. And when a coding assistant writes a great deal of the copy, a checked-in rules file is the only way to make it write in one consistent voice rather than a plausible average one.
- MIT, a single binary, no runtime to install — it will still work in three years.
- Understands Markdown and MDX, so it drops straight into Astro and Starlight (Ed.4 #4) without mangling components.
- Twenty lines of YAML gives you an enforceable house glossary — the cheapest brand consistency available.
- It ships almost no rules of its own: day one is installing or writing a style package, so the first session is configuration rather than results.
- Off-the-shelf styles produce false positives on technical writing until tuned, and an ignored linter is worse than none.
- The repository has moved —
errata-ai/valenow redirects tovale-cli/vale. Ownership changes are worth watching on a tool that a workflow is built around.
