Browser Img Hunter

Browser image capture CLI + MCP server for page-by-page document capture and PDF export.

Use this only for content you are authorized to access and save.

Quick Start

1) Install

npm i -g browser-img-hunter
npx playwright install chromium

2) CLI capture

browser-img-hunter --url "https://example.com/document" --out captures --channel msedge

3) MCP setup in a project (Cursor + Gemini CLI)

cd your-project
browser-img-hunter-init .

This creates a project-local .scp workspace and updates:

• .cursor/mcp.json
• .gemini/settings.json

Features

• One image per detected page number (page-0001.png, page-0002.png, ...)
• Virtualized viewer-safe capture with render readiness checks
• Automatic cookie banner, popup, and promo overlay dismissal before capture
• Improved Scribd-style page detection for text markers like - 1/10 -
• Optional headless capture mode for unattended runs
• Missing/duplicate page verification
• PDF build from captured pages
• Project-local .scp workspace for reusable agent workflows

Commands

| Command | Description | |---|---| | browser-img-hunter | CLI page capture | | browser-img-hunter-init | Initialize .scp + client MCP config | | browser-img-hunter-mcp | Start MCP server (stdio transport) |

CLI Usage

browser-img-hunter --help

Example:

browser-img-hunter --url "https://example.com/document" --out captures --wait-ms 10000 --max-scrolls 1200

Headless example:

browser-img-hunter --url "https://example.com/document" --out captures --auto-start --headless

Common options:

• --url <url> target URL
• --out <directory> output folder
• --profile <directory> browser profile folder
• --max-pages <number> stop after N pages (0 = no limit)
• --max-scrolls <number> max scroll rounds
• --wait-ms <number> initial wait after open
• --delay-ms <number> delay between capture actions
• --ready-timeout-ms <number> page render wait timeout
• --headless run browser without a visible window
• --auto-start start without Enter prompt
• --append keep existing page-*.png files in output

Viewer handling notes:

• The CLI tries to close or hide cookie banners, popups, ad overlays, and promo interstitials before screenshots.
• This is especially useful on Scribd-like viewers where consent banners or recommended content blocks can overlap the page.
• The capture flow re-checks overlays after initial load and again right before page screenshots.

MCP Tools (English Names)

| Tool | Purpose | |---|---| | scp_initialize_project | Create .scp workspace and MCP client config | | capture_document_pages | Capture pages into .scp/captures/<job_name> | | verify_capture_directory | Detect missing/duplicate pages and exact duplicate files | | build_pdf_from_capture | Create PDF from captured page images |

.scp Structure

| Path | Purpose | |---|---| | .scp/captures/ | Capture job outputs | | .scp/profile/ | Reused Playwright browser session | | .scp/logs/ | Runtime logs | | .scp/mcp/ | MCP config templates |

Configuration

| Requirement | Value | |---|---| | Node.js | >= 18 | | Browser engine | Playwright Chromium | | Package name | browser-img-hunter |

Development

npm install
npm run capture:help
npm run mcp:start
npm run scp:init -- .

Publishing

npm run pack:check
npm publish --access public

License

ISC