Contextractor

Contextractor is a free and open-source content extraction tool designed to obtain clean, readable text from any web page by removing navigation elements, advertisements, and boilerplate content. Its primary output is LLM-ready Markdown, which typically reduces token count by 80–90% compared to raw HTML, making it suitable for use in large language model workflows and other text processing tasks. The tool is positioned for users who need to build LLM training datasets, create retrieval-augmented generation (RAG) pipelines, or assemble research corpora.

The extraction engine is based on a Rust port of Trafilatura, which maintains the original heuristic extraction core and incorporates machine learning-based page-type routing for tailored extraction strategies. Crawlee is used for web crawling. 966 F1 score on the ScrapingHub benchmark, and emphasizes compiled Rust performance without requiring Python or a GPU.

Users can access Contextractor via several channels: as a hosted Apify actor, through an npm CLI or library package, or by downloading and running it locally. The tool provides a playground interface for previewing extraction results, adjusting extraction settings, and generating ready-to-run CLI commands, library calls, or Apify actor API calls. Extraction settings allow users to choose between different modes (precision, balanced, recall), and to include or exclude elements such as comments, tables, images, and links. Output formats supported include plain text, JSON with metadata, Markdown, HTML, and original raw HTML.

Contextractor supports extracting a single page, crawling sites, exporting results to files, and managing storage. It is available without login or API key requirements, and users can self-host the tool or use it via the Apify platform, which offers a free usage tier. The source code is available on GitHub, reflecting its open-source nature.