mcp-chrome

Why we included this project

This one targets teams that want an AI agent to work inside the browser people actually use, rather than spinning up a headless instance for every task. It runs as a Chrome extension paired with a Model Context Protocol server, so Claude and other MCP-capable clients can drive real tabs while keeping your existing logins and settings intact. Since the agent operates in the same browser you already use for research and daily work, you skip most of the setup that Playwright-style tooling demands, and the whole thing stays local. It also bundles content extraction and semantic search over open tabs, so a model can find and analyze what is already on screen instead of just clicking around. Teams building agent workflows that need live, stateful browser access are the main audience.