parse-bookmarks

parse-bookmarks

Parse any bookmark export into clean JSON, and convert between formats.

Every browser and read-later app exports bookmarks in a different shape: the ancient Netscape HTML file (Chrome, Firefox, Safari, Edge — with unclosed tags, because 1996), Pocket's CSV, Instapaper's CSV, Raindrop's CSV. If you build anything that touches saved links, you end up writing these parsers yourself. This is that code, written once, tested, with zero dependencies and zero build step.

npm install parse-bookmarks

Parse anything

import { parse } from "parse-bookmarks"

const { format, bookmarks, warnings } = parse(fileContents)
// format   → "netscape-html" | "pocket-csv" | "instapaper-csv" | "raindrop-csv" | "url-list"
// bookmarks→ [{ title, url, addedAt, tags, folder }, ...]

Auto-detection covers:

Source	Format	Notes
Chrome / Firefox / Safari / Edge / Vivaldi export	Netscape bookmark HTML	folders (nested), ADD_DATE, Firefox TAGS
Pocket CSV export	pocket-csv	pipe-separated tags, archive status
Instapaper CSV export	instapaper-csv	folders, timestamps
Raindrop.io CSV export	raindrop-csv	comma tags, ISO dates
A pasted list of URLs	url-list	URL, Title | URL, Title - URL, [Title](URL)

Every parser returns the same unified shape:

interface Bookmark {
  title: string
  url: string
  addedAt: string | null // ISO 8601
  tags: string[]
  folder: string | null  // "Work/Reading"
}

Convert to anything

import { toMarkdown, toCsv, toJson, toNetscapeHtml } from "parse-bookmarks"

toMarkdown(bookmarks)     // - [Title](url), grouped by folder
toCsv(bookmarks)          // title,url,addedAt,tags,folder
toJson(bookmarks)         // pretty JSON
toNetscapeHtml(bookmarks) // HTML file every browser can import

The last one is the escape hatch people ask for the most: Pocket CSV in, browser-importable HTML out — your saved list back in a place that will outlive any single app.

import { parse, toNetscapeHtml } from "parse-bookmarks"
toNetscapeHtml(parse(pocketCsv).bookmarks)

Design constraints

• Zero dependencies, zero build. Plain ES modules + JSDoc + hand-written .d.ts. node --test for tests. Works in Node ≥18 and modern browsers.
• Never throws on dirty rows. Bad rows are skipped and reported in warnings, because real exports are messy.
• No network, no telemetry. It's a parser.

Who makes this

Built by the team behind Burn 451, a read-later app where every save gets 24 hours before it burns. This library is deliberately neutral — use it to build a Burn competitor if you like.

Pairs well with bookmarks-mcp, which uses it to make your live browser bookmarks readable by AI assistants.

License

MIT