Bin-Huang/camoufox-cli

What it solves

It provides AI agents with a way to browse the web while avoiding detection. Many websites use fingerprinting to identify and block automated bots; this tool allows agents to mimic real human users by spoofing hardware and software signatures, maintaining persistent identities, and interacting with pages via a simplified command-line interface.

How it works

The tool acts as a CLI wrapper around the Camoufox browser. It uses a Python daemon that communicates with the browser via Playwright. AI agents send shell commands to the CLI, which are then passed to the daemon to perform actions like navigating to URLs, clicking elements, or taking screenshots. To prevent detection, it spoofs C++-level fingerprints (including canvas, WebGL, and audio) and can save these identities to disk so an agent can return to a site as the same "device."

Who it’s for

Developers building AI agents (using frameworks like Claude Code, Cursor, or OpenClaw) that need to perform web automation, data extraction, or interaction with websites that have strict anti-bot protections.

Highlights

• Anti-Detection: Spoofs canvas, WebGL, audio, screen metrics, and fonts to bypass bot detection.
• Agent-Friendly Interaction: Uses accessibility-tree snapshots with unique element references (@ref) so agents can easily target and click elements.
• Identity Management: Supports persistent fingerprints and cookies, allowing agents to maintain stable sessions across launches.
• Session Isolation: Enables multiple isolated browser processes with different fingerprints and proxies for multi-account management.
• json Output: Provides a --json flag for structured data extraction, making it easier for agents to parse results.