Obsidian Web Clipper + Web2MD: The Complete Clipping Stack for AI Workflows in 2026

The standard answer to "MarkDownload is gone, what do I use to clip webpages to Markdown for Obsidian?" is now Obsidian Web Clipper. It is the right answer.

Here is what that answer leaves out: most of us are not just clipping into Obsidian anymore. We are pasting webpages into ChatGPT, Claude, Cursor, Gemini, and Perplexity dozens of times a day. Obsidian Web Clipper does not help with that moment.

This post is about the workflow that works after you have already installed Obsidian Web Clipper.

The two moments of clipping

When you open the same webpage with the same tool installed, two completely different things might happen next:

Moment A — "I want this in my second brain." You read an article. It is good. You want it filed, tagged, linked to related notes, and searchable in six months. Destination: Obsidian vault.

Moment B — "I need this in this AI prompt right now." You are mid-conversation with Claude. You want it to read the docs page you just opened. Destination: the chat input box, two seconds from now.

Obsidian Web Clipper was built for Moment A. It is the cleanest path from page to vault. The team behind Obsidian shipped it; it has templates, frontmatter, folder targeting, and direct vault writes.

Web2MD was built for Moment B. It is the cleanest path from page to AI prompt. One click converts the page to clean Markdown on your clipboard, or sends it directly into a ChatGPT, Claude, Gemini, DeepSeek, or Perplexity chat in another tab.

They are not in tension. They cover the two moments.

What this looks like in practice

Here is the actual setup on my own machine:

1. Obsidian Web Clipper — pinned in the toolbar. Used roughly 5–10 times a week, always for content I want in my vault.
2. Web2MD — also pinned. Used 30–50 times a week. Almost always followed by Cmd+V into Claude, ChatGPT, or Cursor.

The split is roughly 1:5 — far more "into AI" than "into vault." Most knowledge work in 2026 is on AI surfaces, not in note-taking apps. That ratio is why the companion tool matters.

Where Obsidian Web Clipper wins (and Web2MD doesn't compete)

Honest list of things Obsidian Web Clipper does that Web2MD doesn't:

• Direct vault writes — one click puts a .md file in your vault, in the folder you specify, with the template you defined.
• Templates with frontmatter — {{title}}, {{author}}, {{published}}, {{url}}, {{date}} injected automatically. Web2MD outputs the content; you'd assemble frontmatter yourself.
• Per-site rules — different templates for Wikipedia vs. a personal blog vs. a research paper. Web2MD uses a consistent output format.
• Selection clipping with the right context — clip a paragraph, keep the page metadata.
• Free, open-source, no daily limit — Obsidian Web Clipper has no Pro tier.

If your workflow is Obsidian-first, install Obsidian Web Clipper. Stop reading here unless you also work in AI chats.

Where Web2MD wins (and Obsidian Web Clipper doesn't compete)

Equally honest list:

• Send-to-AI flow — open ChatGPT, Claude, Gemini, DeepSeek, or Perplexity in another tab; Web2MD detects it and offers a "Send to this AI" button. One click, the page is in the prompt.
• Reddit clipping that actually works — Obsidian Web Clipper hits the Reddit DOM, which collapses comments and breaks on Shadow DOM. Web2MD uses Reddit's JSON API path; it captures the full thread including nested replies.
• AI conversation pages — ChatGPT (chatgpt.com/c/), Gemini (gemini.google.com), DeepSeek (chat.deepseek.com), Perplexity (perplexity.ai/search), Claude chat. Web2MD has dedicated extractors that produce ## User / ## Assistant headings. Generic HTML-to-Markdown on these pages produces garbage.
• Chinese platforms — Xiaohongshu (小红书), WeChat public account articles, Zhihu, Bilibili. These platforms aggressively block bot traffic and use SPA rendering. A browser extension is the only category that reliably handles them; Web2MD ships specific extractors. Obsidian Web Clipper's generic extractor returns empty.
• Token counting — before you paste a 30k-character page into Claude, Web2MD shows you the GPT-4 and Claude token estimates. Useful for planning context budget.
• Bulk export for AI conversation migration — queue ChatGPT or Claude conversations, export to one merged .md for migration to another AI or for archiving into Obsidian. This is the feature that actually matters when you're switching between AI providers.
• DOM preprocessing for code, tables, and LaTeX — Wikipedia, arXiv, Stack Overflow conversions strip Highlight.js / Prism / Shiki residue spans, unwrap gist-style line-number tables, expand colspan, and convert KaTeX / MathJax back to TeX. The output is paste-ready in Claude or Cursor; Obsidian Web Clipper preserves the residue.

These are not "extra features" — they are different category of work. Obsidian Web Clipper is a vault writer. Web2MD is a clipboard-to-AI utility.

A concrete example

Here is the same Stack Overflow page through both tools.

Obsidian Web Clipper output (excerpt, saved to vault):

---
title: How to flatten a nested list in Python
source: https://stackoverflow.com/questions/...
published: 2024-03-12
clipped: 2026-05-20
---

# How to flatten a nested list in Python

<span class="hljs-keyword">def</span> <span class="hljs-title function_">flatten</span>(lst):
    <span class="hljs-keyword">return</span> [item <span class="hljs-keyword">for</span> sublist <span class="hljs-keyword">in</span> lst <span class="hljs-keyword">for</span> item <span class="hljs-keyword">in</span> sublist]

The hljs-* spans break the code block. You'd clean them up by hand before pasting into Claude.

Web2MD output (same page, to clipboard):

# How to flatten a nested list in Python

Source: https://stackoverflow.com/questions/...

## Question (score: 247)

I have a nested list like `[[1, 2], [3, 4], [5, 6]]` and I want to flatten it...

## Top answer (score: 892, accepted)

```python
def flatten(lst):
    return [item for sublist in lst for item in sublist]

Or using itertools.chain.from_iterable for performance:

from itertools import chain
list(chain.from_iterable(lst))

Web2MD strips the highlighting markup, preserves the code block, and adds vote scores as context. Paste into Claude → Claude reads it correctly the first time.

## The combined workflow

Here is what I'd recommend if you do both kinds of work:

1. Install **Obsidian Web Clipper** ([from obsidian.md/clipper](https://obsidian.md/clipper)). Configure your default folder, your filename template, and one or two per-site templates. Pin to toolbar.
2. Install **Web2MD** ([from web2md.org](https://web2md.org?utm_source=blog&utm_campaign=obsidian-companion)). No setup required. Pin to toolbar.
3. **Decide by destination, not by tool.**
   - If the page is going into your vault → Obsidian Web Clipper.
   - If the page is going into an AI chat → Web2MD.
   - If both — clip with Obsidian Web Clipper first (it writes the canonical copy), then use Web2MD to send to AI.

This is not zero-sum. The two tools have different jobs.

## What MarkDownload users actually lost

MarkDownload users were a mix. Some were Obsidian-first (Obsidian Web Clipper is now your tool). Some were "I want clean Markdown on my clipboard for whatever's next" (Web2MD is now your tool).

If you've been using Obsidian Web Clipper and felt something was missing for AI work — that's what this post is about. Add Web2MD as the second tool. The combined stack covers what MarkDownload used to do plus the AI workflows MarkDownload was built before.

## Internal references

- [Best Web Clipper in 2026 — After MarkDownload's Removal and Pocket's Shutdown](/blog/web-clipper-comparison-2026-after-markdownload-pocket)
- [Best MarkDownload Alternative for Obsidian](/blog/markdownload-alternative-obsidian-2026)
- [Obsidian Web Clipper vs Web2MD: an honest comparison](/blog/obsidian-web-clipper-vs-web2md)
- [Web2MD vs MarkDownload: feature-by-feature](/alternatives/markdownload)

## Install

[Web2MD on the Chrome Web Store →](https://web2md.org?utm_source=blog&utm_campaign=obsidian-companion)

3 conversions per day on the free tier, no signup required. Pro is $9/month if you outgrow it.