linkpeek vs metascraper
Compare linkpeek with metascraper: a focused preview library against a rule framework. Speed, install size, fetching, and how to migrate.
metascraper is a rule framework maintained by Microlink. You hand it HTML and a URL, and each rule package you install (title, image, author, date, logo, and provider packages such as YouTube, Amazon or Spotify) contributes values with its own fallbacks. It does not fetch. linkpeek is a smaller preview library: it does the fetch itself, with the network checks, reads the page head and returns one fixed, typed result.
Measured side by side
| Metric | linkpeek 3.0.0 | metascraper 5.58.6 |
|---|---|---|
| 500 kB page, median | 2.76 ms | 62.6 ms |
| Small pages, mean of medians | 1.25 ms | 2.71 ms |
| Clean install size | 1.3 MB | 81.1 MB |
| Packages installed | 7 | 128 |
metascraper does not fetch, so it was measured as native fetch plus metascraper with seven rule packages: author, date, description, image, publisher, title and url. The install size counts the same seven packages. Its large-page times varied widely between runs. See the benchmarks for the method and every package.
Choose metascraper when
- You need provider-specific rules, such as
metascraper-youtube,metascraper-amazonormetascraper-spotify, each tuned to how that site marks up its pages. - You want to write your own rules or change which source wins for a field. linkpeek's fallback order is fixed.
- You fetch HTML your own way, for example through a headless browser for pages that render their tags with JavaScript, and only need extraction on top.
Choose linkpeek when
- You want fetching included. With metascraper, redirect validation, address policy, timeouts, byte limits and credentials belong to your own fetch. linkpeek validates the URL and every redirect, blocks private IP literals, caps the bytes it reads and times out by default.
- You run on Cloudflare Workers, Vercel Edge, Deno or Bun. metascraper's maintainers describe browser and edge use as not what it is oriented for; linkpeek uses only platform
fetch, Web Streams andTextDecoder. - Install size matters: linkpeek has one runtime dependency and no rule packages to choose.
- You need preview-card fields, not a framework: title, description, image, site name, favicon and canonical URL, typed as a value or
null.
Migrate metascraper to linkpeek
npm install linkpeek// metascraper
import createMetascraper from "metascraper";
import description from "metascraper-description";
import image from "metascraper-image";
import publisher from "metascraper-publisher";
import title from "metascraper-title";
const metascraper = createMetascraper([description(), image(), publisher(), title()]);
const html = await (await fetch(url)).text();
const meta = await metascraper({ html, url });
// meta.title, meta.description, meta.image, meta.publisher
// linkpeek: fetch included
import { preview } from "linkpeek";
const result = await preview(url);
// result.title, result.description, result.image, result.siteNameIf you keep your own fetch, swap only the extraction step: parseHTML(html, url) is synchronous and returns the same result shape as preview(), with statusCode set to 0. Your fetch then still owns redirect checks, timeouts and byte limits.
| metascraper field | linkpeek field |
|---|---|
title | title |
description | description |
image | image |
url | canonicalUrl |
publisher | siteName |
author | author |
date | publishedDate |
lang | lang |
linkpeek has no equivalent of metascraper-logo; favicon is the declared icon or the origin's /favicon.ico. HTTP error pages resolve with their statusCode; check it before caching. The API reference lists every field and error code, and the security model explains the opt-in resolveHost check and the DNS limit to cover with egress rules.
New to linkpeek? The quick start covers install, presets and where it runs.