ScrapLLM - Web to Markdown for AI by dekrezz

ScrapLLM turns whatever is on screen into clean Markdown, ready to paste into a model. One click for a page, one for a highlighted fragment, one for a whole conversation.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

WHAT IT COPIES

• Convert & Copy — the page as Markdown. The article body by default, the full page on demand.
• Copy Selection — only what you highlighted, with its source and line range. Code is fenced with its language.
• Copy Chat — an LLM conversation with roles and timestamps: the whole thread, or the last few exchanges.
• Download Markdown — the same output as a .md file.
• Multi-tab — several open tabs at once, merged into one file or zipped separately.
• Research — ask a question, get one Markdown file: it searches, reads each source, and downloads the result.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

SITES IT UNDERSTANDS

Most extractors stop at what a reader-mode algorithm can see. These go further, because on these sites the part worth copying is the part that gets dropped:

• Reddit — the post with its full comment tree, including replies folded behind "more replies". Deleted comments are removed and their surviving answers keep their place.
• GitHub — a repository as its README in the original Markdown plus the complete file tree; a discussion as its post and every reply, with authors and roles.
• YouTube — the description in full, not the two lines the page shows, plus the comment thread with likes and dates.
• Telegram — a channel, group, forum topic, bot or private chat as a transcript, with reply context and reactions.
• Discord — the open channel as a transcript with authors, timestamps and attachments.
• X — threads with replies, quoted posts and engagement counts.
• AI chats — ChatGPT, Claude, Gemini, Grok, Perplexity, DeepSeek, Copilot, Cursor, and self-hosted front-ends.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

PRIVACY

Conversion happens entirely in your browser. Page content is never uploaded, there is no account, and there is no telemetry.

The optional token counter downloads tokenizer data from tiktoken.pages.dev once and caches it. Research is the single feature that contacts a third party: it sends your question to DuckDuckGo to find sources, then reads each source itself — without your cookies. Everything else works offline.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

INTERFACE

Spring-driven motion you can interrupt mid-flight. Press feedback lands as you press, not as you release. Light and dark themes, translucent materials, and full support for reduced motion, reduced transparency and increased contrast.