Turn any PDF in Chrome into
clean, copyable text.

Extracts structured Markdown from PDFs that Chrome's native viewer mangles. Runs 100% locally via WebAssembly. No uploads. No servers. No tracking.

Install from Chrome Web Store Free. Open source engine. 1.1 MB total.

How it works

  1. Open any PDF in Chrome. Navigate to a .pdf URL. Chrome's native viewer loads as usual. The PDF to Text icon shows a blue badge while it pre-fetches the bytes in the background.
  2. Click "Open in PDF to Text." The popup offers one-click extraction. No auto-redirect, no overriding your preferred viewer. Your choice, every time.
  3. Copy the clean text. The viewer shows extracted text with headings, paragraphs, and a Table of Contents auto-detected from font sizes. Copy as plain text, basic Markdown, or structured Markdown with YAML frontmatter and token counts.

What you get

Privacy
Everything runs locally. The PDF bytes go from your browser to a WebAssembly module and back. Zero network calls to any server we control. Zero analytics. Zero telemetry.
Rescued glyphs
A 7-level fallback cascade recovers text from PDFs with broken Unicode mappings, missing ToUnicode CMaps, subset-embedded fonts, and ligature encoding bugs. The glyphs Chrome would show as gibberish come out as real words.
Agent-ready
Browser automation tools (Playwright, Puppeteer, Claude in Chrome) can read window.__glyph.markdown.structured directly. No API key. No rate limit. No server.

Use with Claude Code

If you use Claude Code, you can add PDF to Text as a skill so Claude can invoke it automatically when you ask about PDFs. 

/plugin marketplace add JordanCoin/glyph-api
/plugin install glyph-api@glyph-api

Requires collaborator access to the private repo. The extension itself works for everyone via the Chrome Web Store.

Install

Chrome Web Store →

Link will be live once the extension passes Chrome Web Store review. Free. No account required. 1.1 MB download.

Privacy

Your PDFs never leave your device. PDF to Text runs entirely in your browser using a local engine. We collect only anonymous performance stats (extraction time, page count) to improve the tool. No PDF content, no URLs, no identity.

Read the full privacy policy →