Skip to content

Give your AI a browser that can reach protected websites

Cloak Biz Scraper is an open-source browser service you deploy yourself. It gives Claude or ChatGPT a remote MCP connection to a patched Chromium browser running through your residential proxy.

Set it up once, then use it for a scheduled business-listing watch or any authorized public page that blocks an ordinary AI browser.

Set up the scraper for AI View the open-source project

Example morning review with scraped listings, triage decisions, and archive status

One workflow you can build: a morning review queue. This illustrative sample uses fictional listings; your queue will contain the businesses found by your saved searches.

First, connect the browser to your AI

The shared setup guide walks through the complete foundation:

  1. deploy Cloak Biz Scraper on Railway;
  2. get the CloakBrowser Pro key through its checkout email;
  3. generate Evomi proxy credentials and split the copied string into the four required fields;
  4. connect the scraper's /mcp URL to Claude or ChatGPT with OAuth; and
  5. prove that the AI can call create_instance and agent_browser in a harmless test.

Then choose a workflow

Wake up to businesses worth reviewing

The listing tutorial adds Notion, reusable search-result URLs, objective triage rules, page archiving, and a daily Claude or ChatGPT Work schedule. Each morning the agent calls scrape_listings, filters new listings, calls archive_page for anything it keeps or cannot decide from the summary, and leaves a review queue.

Build the daily listing watch

Browse a page with anti-bot detection

The protected-site guide provides copy-and-paste prompts that tell the AI exactly which MCP and tools to use. It covers one-page reads, multi-step searches, custom instructions, archiving to Notion, CAPTCHAs, and a single bounded proxy rotation.

Open the protected-site prompt guide

A live CloakBrowser instance the AI can operate

Watch and maintain it

  • Monitor and take control explains the live preview, browser takeover, task history, saved evidence, and human login or CAPTCHA handoff.
  • Advanced controls explains proxy-session rotation, concurrent browser limits, the task versus interactive-session mix, durable profiles, and disk cleanup.
  • Daily agent runbook is the exact operating procedure to copy into Notion for the scheduled listing workflow.

Verified against the real workflow

The guides were checked against the running Pro browser and residential proxy. A one-page BizBuySell sweep returned 50 listings with sync=false. The archive tool then saved a synthetic page under Source Content, and Claude read that archived body back through its Notion connector without editing it.

Claude reading content saved by archive_page through Notion

See the verification record for the test boundary and provider documentation used before publication.