PagePress
Say it on the page.
Capture the whole web page. Mark up exactly what you mean. Send it to a teammate, a client, or the AI agent writing your code.
Works in Chrome, Brave and Firefox. No account. Nothing leaves your browser.
You know exactly what’s wrong with the page.
“Scroll down. No, further.”
“The other Learn More button.”
“It’s somewhere below the fold.”
Stop describing the page. Show it.
Capture
The whole page. One click.
Your screenshot tool stops at the edge of the window. PagePress keeps going.
Every screen, stitched into one.
No seams. No repeated headers.
It scrolls the page for you, captures every screen, and joins them into a single image. Sticky headers appear once, at the top, where they belong.
Export
Know the size
before you save.
Every format is encoded for real before you choose. The number under the button is the file you’ll get. No surprise eleven-megabyte attachments.
One real capture of a long page, measured in each format.
Every pixel.
The default. Keeps everything your display captured, for archives, print, and anything you might crop into later.
A quarter of the size.
Resampled to the page’s own pixels. Small enough for a ticket or a chat, and it reaches twice as far down very long pages.
Edit
Point. Number.
Redact.
A screenshot rarely speaks for itself. Press Edit and your capture opens in its own tab, ready to make your point for you.
Arrows, boxes, pen and text.
Drawn at the capture’s full resolution, so they stay crisp in the file you send.
Steps that count themselves.
Delete step two of five and the rest renumber. 1, 2, 3, 4. Never a gap.
Redaction that really redacts.
Solid is the default because it destroys the pixels underneath. Blur and pixelate are there when you want the shape to show.
Crop without regret.
Crops and rotations never touch the original. Change your mind an hour later and widen it back out.
Undo, all the way back.
Every mark, every crop, every rotation. And redo, for when you were right the first time.
Copy it, or save it.
Straight to your clipboard, or downloaded as PNG, JPEG, WebP or a PDF sized for the screen or for the printer.
AI Markup New
Point at it.
Your agent fixes it.
Building a site with Claude Code or another AI coding agent? Stop typing out which button you mean. Mark up the live page itself: comment on any text, strike what should go, point at anything. Then hand your agent the whole review in one paste.
It speaks your agent’s language.
Your agent edits source files, and nothing in a codebase is organised by pixel. So every note quotes the words on the page it’s about. Your agent searches for them, finds the exact line, and makes the change.
No screenshots to squint at. No “the blue button near the top.” Just a clear list of what to change and where.
# Page review **Title:** "Northwind Roasters" **Page:** `https://northwindroasters.com/` **Marks:** 3 Each item names an element by the text it displays. **Locate it by searching the source for the quoted text, not by the CSS selector.** The selector describes the rendered DOM, which a build step may have generated; the text is what actually appears in the template. Before editing, confirm the quote matches exactly one place. If it matches several, use the **Within** line to choose. If it is still ambiguous, say so and ask. Do not guess. ### 1. Comment **Element:** `h1` **Reads:** "Coffee with a place of origin, not a marketing story." **Within:** `div.hero → div.wrap` **Instruction:** say who it is for, not just what it is ### 2. Remove **Element:** `p.eyebrow` **Reads:** "Harvest 2026 · Now shipping" **Within:** `div.hero → div.wrap` ### 3. Points at **Element:** `a.btn.btn-p` **Reads:** "Shop single origin" **Within:** `div.hero → div.btns` **Instruction:** this should be the only button here
Anchored to words, not pixels.
Every mark leads with the text it sits on, the one thing that survives the trip from the page back to your source.
Marks stay put.
Images load, menus open, the layout shifts. Your notes stay on the words they belong to.
Prefer a picture?
Press Review and PagePress captures the page with your marks drawn on it, with the prompt one click away.
Works with Claude Code, Cursor, Copilot, and any agent that can read your repository.
Privacy
What happens on your page
stays on your page.
PagePress has no server to send anything to. Every capture is built inside your browser and stays on your device until you save it or copy it.
No server.
Captures, edits and prompts are made on your machine, and nowhere else.
No account.
Nothing to sign up for. Install it and it works.
No tracking.
The extension reports nothing. No analytics, no telemetry, no idea who you are.
No access until you click.
It installs with no permission to read any site. It sees a page only when you ask it to.
Tech specs
- Browsers
- Chrome, Brave and Firefox on desktop
- Capture
-
Full page, stitched from scrolling screenshots, with sticky headers shown once
Follows pages that scroll inside a panel, and content in embedded frames, asking for access to a frame only at the moment it’s needed
- Formats
-
PNG, JPEG, WebP and PDF, each with its real file size shown before you save
JPEG and WebP quality from 40 to 100. PDF as one continuous page, or Letter and A4 pages for printing
- Scale
-
Retina keeps every pixel (default). Standard is roughly a quarter of the file size.
- Editor
- Arrows, boxes, pen, text, numbered steps, redaction (solid, blur, pixelate), crop, rotate, undo and redo, in eight colours
- AI Markup
-
Comment, remove, arrows, boxes, pen, notes, steps and redaction on the live page
Copy prompt exports the review as markdown. Review captures the marked-up page into the editor.
- Shortcut
- ⌘⇧Y on Mac, Ctrl+Shift+Y on Windows and Linux, to capture and download in one step
- Permissions
-
activeTab,scripting,downloads,storage,clipboardWrite, andoffscreenon ChromeNo host permissions at install. No debugger permission, so no “being debugged” banner across your browser.
- Version
- v1.4.0. The popup tells you once what changed after each update.
- Price
- Free
Where it stops
- Long pages take a few seconds. Browsers allow about two screenshots a second, and PagePress paces itself just under that.
- Very tall pages get cut off. A canvas can’t exceed 32,767 pixels on a side: about 16,000 pixels of page at Retina on a 2× display, or 32,000 at Standard. PagePress tells you when it stops rather than handing you a partial image quietly.
- Browser pages can’t be captured. Settings pages, the extension stores and other extensions’ pages are off limits to every extension.
- Frames inside frames aren’t followed. Only the outermost embedded document is captured.
- Two big scrolling areas on one page. It captures the bigger one.
Say it on the page.
Free for Chrome, Brave and Firefox.