html-to-markdown
High performance and CommonMark compliant HTML to Markdown converter.
Parse messy real-world HTML into clean CommonMark across 16 different programming languages.
This tool transforms unclosed tags, malformed entities, and nested tables into clean CommonMark or Djot. It suits developers building data pipelines or RAG systems that ingest raw web content.
What it is
It uses a tiered dispatch architecture combining byte scanners, DOM walkers, and html5ever repair to process real-world HTML at 19 to 116 MB/s. Alongside text conversion, it extracts structured metadata like Open Graph, Twitter cards, and JSON-LD in a single pass.
Who it's for
It suits software engineers and data professionals processing large corpuses of web pages across diverse language runtimes. It does not suit end-users looking for a graphical desktop application or a no-code web clipping service.
Key features
16 Language Bindings
Call the Rust core directly from Python, Node.js, Go, Java, PHP, and 11 other languages.
Tiered HTML Repair
Automatically handles unclosed tags, CDATA, and broken entities without losing content.
Metadata Extraction
Extracts Open Graph, Twitter, and JSON-LD metadata during the conversion pass.
Djot Output Support
Switch output formatting from standard CommonMark to Djot via configuration.
Up and running in 3 steps
Our honest verdict
What's great
- Delivers byte-equal output across 16 different language bindings using a single Rust core
- Extracts structured page metadata and tables simultaneously with text conversion
- Processes large corpuses at speeds between 19 and 116 MB/s
- Includes agent skills and plugins for AI coding assistants like Claude Code and Cursor
Where it falls short
- Some language bindings require specific native installation steps like PIE for PHP
- Requires integrating package dependencies into your existing codebase
- Offers a programmatic API rather than an out-of-the-box graphical user interface
Pricing & license
Free under its open source license, with the only cost being your own compute resources.
Open remote AI-training roles
873 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
What output formats does it support?
It outputs CommonMark-compliant Markdown by default, with an option to emit Djot format instead.
How does it handle malformed HTML?
It uses a tiered dispatch system featuring a byte scanner, DOM walker, and html5ever repair to handle unclosed tags and mixed encodings automatically.
