Fetch a web article, scrape it to Markdown, then use Apple's on-device Natural Language framework to color-tag every word by part of speech — a fun, instant linguistic breakdown with zero API calls to a language model for the tagging itself.
Analyzing the grammatical structure of a piece of writing (nouns vs. verbs vs. adjectives, etc.) usually requires a dedicated NLP tool or library — not something you'd casually ask a chatbot for, since it needs actual linguistic parsing, not just "reading comprehension."
Minis combines browser-use (or a scraping tool) to fetch and convert a web article to Markdown, then pipes the text through Apple's native NaturalLanguage framework (via the apple-nlp CLI) to tag every token with its part of speech. It renders the result as an interactive, color-coded HTML page — nouns in blue, verbs in red, adjectives in purple, etc. — with a hoverable tag legend and per-word counts.
None (built-in). Uses the native apple-nlp CLI plus a Markdown-scraping tool (the original review used Firecrawl; web-content-extractor from MinisSkills works as a local alternative) and the built-in WKWebView for rendering.
- Give Minis a URL to an article (or paste the text directly).
- Ask it to scrape it to Markdown and tag parts of speech with Apple NLP.
- Minis renders an interactive HTML report you can scroll through, with a legend showing counts per part-of-speech category.
Fetch my most recent article from [site], scrape it as Markdown, tag its different parts
of speech with Apple NLP, and present it nicely.
An HTML page titled with the article's headline, byline, and total tagged-token count, followed by a legend (Noun / Verb / Adjective / Adverb / Pronoun / Determiner / Preposition / Conjunction / Interjection / Number / Particle / Other, each with a running count), then the full article text rendered with every word color-highlighted and hoverable for its exact tag.
- A way to fetch/scrape article text (browser_use, a scraping skill, or pasted text directly)
- No API key needed for the NLP tagging itself — it's fully on-device via Apple's framework
- Great demo of combining browser-use + a native Apple framework + HTML artifact rendering in one request.
- Try it on your own writing to spot overused word patterns (e.g. adjective-heavy sentences).
- Works fully offline for the tagging step once the text is in front of Minis — only the initial fetch needs network access.
Sourced from: Open Minis Is the iOS Agent I Wish Siri AI Could Be by Federico Viticci, MacStories (2026-07-20).
2026-07