Web clipper

Web clipper

track-fetch-web clips a web page into your vault: it fetches the page, extracts the readable main content (dropping navigation, sidebars, and other page furniture), and converts it to Markdown so a page becomes a note in one pipeline. Like all track-fetch-* tools it is a separate binary — track itself never talks to the network.

Back to track. See CLI for note commands and Charts for the Canonical Data Model that fetch tools share.

Clipping a page into a note

The --note flag prints a ready-to-pipe Markdown note body — a provenance line, the lead image, then the content — which track new reads from stdin:

track-fetch-web --note https://example.com/essays/growing-tomatoes \
  | track new --title "Growing tomatoes"

The new note starts like this:

[Source](https://example.com/essays/growing-tomatoes) — clipped 2026-07-10

![](https://example.com/images/tomatoes-lead.jpg)

Container gardening rewards small, steady adjustments. A pot dries faster than a bed…

From there it is an ordinary note: add [[Title]] links, tags, and your own commentary. track open and track append work the same way if you prefer to clip into an existing note.

What gets extracted

• Title — from og:title, the <title> element, or the first heading. A leading heading that repeats the title is dropped from the body so it is not duplicated.
• Timestamp — the page's declared publication time (article:published_time and friends), falling back to the fetch time.
• Lead image — og:image, or the first image in the content.
• Main content — a compact readability heuristic: the page's own <article>/<main> container when it carries real text, otherwise the element with the most paragraph text and the fewest links. Headings, lists, quotes, code blocks, tables, links, and images are converted to Markdown; scripts, navigation, footers, and ad/share/comment furniture are removed.

Pages rendered entirely by JavaScript have no readable HTML to extract; the clip degrades to the page metadata.

Canonical JSONL output

Without --note, the tool follows the track-fetch-* contract: it emits one Canonical Data Model event record (JSONL) with the Markdown content and lead image as extra fields, so clips can also feed Charts — a reading-log timeline, for example.

track-fetch-web https://example.com/essays/growing-tomatoes

{"version":1,"time":"2026-05-04T09:30:00Z","title":"Growing tomatoes","url":"https://example.com/essays/growing-tomatoes","image":"https://example.com/images/tomatoes-lead.jpg","markdown":"Container gardening rewards…"}

--out <file> writes the record to a file (conventionally the vault's data/ directory) and prints a JSON summary instead, matching the track CLI's result style.

Flags

Fetches are guarded against SSRF: private, loopback, and link-local addresses are refused, so a clipped URL can never probe your local network. To clip a page from a local or internal site, save it to a file first and pass the path.