Anthropic Launches Claude Chrome Extension With Full Browser Automation and Multi-Layer Security Controls
Summary
Anthropic launches a powerful Claude Chrome extension featuring full browser automation with 21 tools, dual operating modes for speed and capability, and a robust multi-layered security system that classifies domains and enforces permission controls to prevent cross-domain attacks.
Key Points
- Anthropic's Claude for Chrome extension (v1.0.56) operates as a Manifest V3 side panel AI agent that runs two distinct modes: a Standard Mode using a full agentic tool-use loop with claude-sonnet-4-5 and a Quick Mode that bypasses tool definitions in favor of a compact single-letter command language for low-latency execution.
- The extension provides 21 tools enabling full browser automation including mouse and keyboard control via Chrome DevTools Protocol, accessibility tree reading, JavaScript execution, network request monitoring, file uploads, GIF recording, and MCP integration with Claude Desktop and Claude Code via native messaging.
- A multi-layered permission system governs every tool execution, classifying domains via Anthropic's API into safety categories, supporting three permission modes including plan-based pre-authorization, and enforcing URL verification before mutating actions to prevent cross-domain attacks during automation sessions.