antibrow/anti-detect-browser-skills

Anti-Detect Browser - Claude Code Skills

Agent Skills for driving a browser that presents one coherent real device instead of a headless build: persistent isolated profiles, kernel-level fingerprints, and a per-profile proxy whose exit IP sets timezone and WebRTC. They teach an agent how to launch and manage AntiBrow browsers, both by writing code against the SDK (JavaScript or Python) and by letting the agent drive the browser directly via MCP.

Typical uses: QA and bot-detection testing against your own site, checking your own ads and pricing from another region, scraping public data, giving an agent a browser that stays logged in, and keeping identities you own or are authorized to operate in genuinely separate profiles. See Acceptable use.

Skills in this repo

• multi-account-isolation - verifying that profiles are actually isolated rather than assuming it: ten concrete checks (timezone vs exit IP, WebRTC, canvas stability across relaunches, duplicate personas or addresses across a fleet), which detection suites to run, what the runtime does with your credentials, and which layers browser isolation cannot cover at all. Start here if the question is "is my setup actually holding?".
• anti-detect-browser - SDK (npm anti-detect-browser + PyPI antibrow), profiles, fingerprints, proxies, kernel updates, Docker, and the REST API. Use this to write custom scraping, multi-account, or automation scripts.
• browser-mcp-agent - MCP server mode. Use this to let an AI agent (Claude, GPT, etc.) launch and control the browser itself via tool calls, with no code to write.

Install

As a Claude Code plugin (recommended)

Installs both skills together and keeps them updatable via /plugin update:

/plugin marketplace add antibrow/anti-detect-browser-skills
/plugin install anti-detect-browser-skills@antibrow

As individual skills

Works with Claude Code and any o