---
source_url: "https://www.contextractor.com/about/?utm_source=openai"
title: "About Contextractor, the web content extraction tool 🧰"
mirrored_at: 2026-08-28T01:00:58.891Z
host: www.contextractor.com
cited_in_42a: true
mirror_canonical: "https://index.42a.ai/www.contextractor.com/about/index__q__utm_source_openai"
---

> **Original source:** https://www.contextractor.com/about/?utm_source=openai

![](https://www.contextractor.com/_next/static/media/cover-mini.39_i7c1bmz5_b.svg)

## What is Contextractor?

Contextractor is an **open-source** web content extraction tool aimed at getting the readable article out of a page, not the whole DOM. The extraction runs on the [Rust port](https://github.com/Murrough-Foley/rs-trafilatura) of [Trafilatura](https://www.contextractor.com/trafilatura/), and the crawling is handled by [Crawlee](https://crawlee.dev/), a TypeScript crawler that can drive [Playwright](https://playwright.dev/) (adaptive by default, or Firefox/Chromium) or fetch pages over plain HTTP with Cheerio when no browser is needed. The tool strips away navigation, ads, and boilerplate, leaving just the text you need. The website at [contextractor.com](https://www.contextractor.com/) serves as a playground where you can enter a URL, configure extraction settings, and preview results.

## Why the Rust port of Trafilatura?

On the [Scrapinghub article set](https://github.com/scrapinghub/article-extraction-benchmark) (181 articles), the [Rust port](https://github.com/Murrough-Foley/rs-trafilatura) scores an F1 of 0.966 (precision 0.942, recall 0.991) — ahead of go-trafilatura at 0.960 and the original Python [Trafilatura](https://www.contextractor.com/trafilatura/) at 0.958[1](#user-content-fn-rs-trafilatura-bench). Trafilatura's hybrid extraction approach is the engine's foundation, and the Rust port keeps that lead while adding auxiliary ML page-type classification and confidence scoring on top of the heuristic extraction core (the extraction step itself stays ML-free). The Rust port is exposed to TypeScript through a napi-rs binding, so every surface — the npm CLI and library, the Apify Actor, and the playground — gets the same extraction quality with no Python runtime involved.

## Token-efficient output for LLMs

Because Contextractor returns just the readable content — navigation, ads, scripts, and boilerplate stripped — its Markdown output typically runs **80–90% fewer tokens than the raw HTML**[2](#user-content-fn-token-reduction). For an LLM or RAG pipeline that is the practical payoff: the same page costs a fraction of the input tokens to process.

## Features

-   **Playground** — Enter a URL and preview extraction results at [contextractor.com](https://www.contextractor.com/)
-   **High accuracy extraction** — the [Rust port](https://github.com/Murrough-Foley/rs-trafilatura) of [Trafilatura](https://www.contextractor.com/trafilatura/) and its hybrid extraction approach for the best balance of precision and recall
-   **Multiple output formats** — [Plain text](https://www.contextractor.com/text/), [Markdown](https://www.contextractor.com/markdown/), [JSON](https://www.contextractor.com/json/), and [HTML](https://www.contextractor.com/html/); the playground's Original raw HTML checkbox additionally returns the raw page source
-   **Metadata extraction** — Automatically extracts title, author, publication date, and more
-   **Free and open-source** — No registration required, source code on [GitHub](https://github.com/contextractor/contextractor)

## Use cases

-   **Build LLM training datasets** — Extract clean text from web pages for machine learning and AI applications
-   **Feed content into RAG pipelines** — Get clean, structured content for retrieval-augmented generation
-   **Research and academic text extraction** — Collect article content for analysis without boilerplate noise
-   **Content monitoring** — Re-extract clean page content on a schedule to feed your own change-detection or diffing pipeline

## Apify Actor

For automated and large-scale extraction, use the [Contextractor Apify Actor](https://apify.com/glueo/contextractor?fpr=glueo). It supports JavaScript rendering with Playwright, link crawling across sites, and configurable extraction modes.

## The company behind Contextractor

_Contextractor_ is operated by [Glueo, s.r.o.](https://www.glueo.com/?utm_source=contextractor&utm_medium=own-referral&utm_campaign=about&utm_content=text), a Prague-based software company that runs its own online services, such as _Contextractor_, and provides custom software development.

1.  rs-trafilatura: [README — benchmarks](https://github.com/Murrough-Foley/rs-trafilatura#scrapinghub-article-extraction-benchmark). Retrieved May 31, 2026 [↩](#user-content-fnref-rs-trafilatura-bench)
    
2.  Measured by tokenizing the raw fetched HTML against Contextractor's Markdown output with the o200k\_base (GPT-4o) tokenizer across a 28-page corpus (news, documentation, encyclopedia, articles, blog, academic, and legal pages): median 90%, corpus-wide 82% fewer tokens. The exact saving varies with how much boilerplate a page carries. Retrieved June 18, 2026 [↩](#user-content-fnref-token-reduction)
    

Updated: July 5, 2026