Any web content.
Clean markdown.
One click.
Markforge is a Chrome extension with dedicated parsers for 17 platforms: tweets, Reddit threads, AI conversations, GitHub repos, Stack Overflow, and more. Not a generic HTML converter. Every parser understands the page structure and produces markdown that actually preserves what matters.
Free. No account required.
Built for Obsidian users, prompt engineers, and anyone who thinks in markdown.
Web clipping is broken
Generic conversion kills the content
You clip a tweet thread and get a wall of mangled divs. You clip a Reddit discussion and it stops at the first link. You clip an article and the code blocks lose their formatting. Generic HTML-to-markdown doesn't understand what it's converting.
The good tools are dead or locked in
MarkDownload never moved to Manifest V3, so Chrome disabled it and its store listing is gone. Obsidian Web Clipper only works with Obsidian and doesn't download images. Readwise costs $10/month for a full reading platform when all you want is the markdown.
Your LLM is eating garbage
Feed a model the HTML of a web page and most of what it reads is markup. On MDN's reference page for Array.prototype.map(), the HTML source weighs 48,153 tokens. The same page converted by Markforge: 3,373. That is 14 times more context than the content needs.
Markforge fixes all three.
Three steps. That's it.
Click the extension
On any web page, click the Markforge icon (or press Alt+M). It detects the platform automatically. Want just one section? Use Clip Selection to visually pick any element on the page.
Edit and choose your format
Preview and edit the title and content before saving. Copy to clipboard, download as .md with images, or save directly to a custom folder. No "Save As" dialogs.
Use it anywhere
Clean markdown with customizable YAML frontmatter, ready for Obsidian (with wikilink support), Logseq, Notion, your RAG pipeline, or any markdown editor.
More than just a converter
Full control over what gets extracted, how it's formatted, and where it's saved. No compromises.
Clip Selection
Don't need the full page? Enter visual picking mode: hover to highlight any element, click to clip just that section.
Smart image download
Images are fetched locally with a 3-step fallback pipeline. Batch processing with real-time progress. PNG, JPEG, WebP, SVG, AVIF and more.
Custom download folder
Pick any folder on your machine. Markforge writes files there directly, with no download manager and no "Save As" dialog. Persists across sessions.
Keyboard shortcuts
Alt+M opens the popup on any page. The shortcut can be changed in the Chrome extension settings.
Content filtering
Strips tracking parameters (utm, fbclid, gclid...), removes elements by CSS selector, filters images by size. Clean output, every time.
Editable before saving
Both the filename and markdown body are fully editable in the popup. What you see is what you get: the file matches the preview exactly.
Not a generic converter. A platform-aware one.
Every platform structures content differently. A tweet isn't an article. A Reddit thread isn't a blog post. Markforge uses a strategy pattern: 20 dedicated extractors that understand each platform's DOM and pull out exactly what matters.
Twitter / X
Tweet threads and articles, author info, quoted tweets, optional replies
Posts with nested comments and their authors, on old and new Reddit
ChatGPT / Claude / Gemini / Grok
Full AI conversations with User and Assistant turns and fenced code blocks
GitHub
READMEs as markdown, source files saved with their own extension
Stack Overflow
Questions + all answers with scores and code blocks
arXiv
Papers with title, authors, abstract, structured sections
Wikipedia
Articles with their section hierarchy, minus infoboxes and footnotes
Substack / Medium
Newsletter and blog articles, clean content extraction
Pulse articles and profile posts with author info
Hacker News
Posts + comment tree with scores and metadata
Gmail
Emails with optional quoted replies preserved
ResearchGate
Publications with academic metadata
Google Search
Search results as a list: title, site and snippet
Any web page
Main content extraction with metadata from the page
Every parser outputs consistent YAML frontmatter: title, author, date, source and URL, when the page provides them.
Stop wasting tokens on HTML
CONTAINS
CONTAINS
Just the contentMeasured on October 14, 2026 on MDN's reference page for Array.prototype.map(): the page's HTML source against the markdown produced by Markforge's parser, counted with the same GPT tokenizer as the popup. Savings depend on the page: 25% on a plain HTML essay, 98% on a GitHub repository page.
Markforge keeps the content and leaves out what your model doesn't need: navigation, scripts, sidebars, buttons. What's left is structured markdown, with its token count shown before you paste.
How Markforge compares
Platform-specific parsers
AI conversation export
Clip selection mode
Token counter
Image download
Works with any markdown app
Maintained
Price
Browser extension
Built by a markdown nerd, for markdown nerds
Markforge started as a personal tool. I was spending 15-20 minutes a day reformatting web content for my Obsidian vault and my LLM workflows. I tried every clipper out there. None of them understood that a tweet needs a different parser than an article. So I built one that does.
Now I use it 30+ times a day. Every article I research, every tweet thread I want to keep, every Claude conversation worth saving: one click, clean markdown, in my vault.
Frequently asked questions
Your web content deserves better than copy-paste.
Install Markforge and start converting web content to clean markdown in seconds.
Free forever. No account, no tracking, no strings attached.