unagi-cq/cdp-bridge-mcp — explained in plain English
Analysis updated 2026-05-18
Let an AI coding agent read content from sites you are already logged into.
Scrape data from authenticated pages without exporting cookies or credentials.
Automate interactions in your real running browser instead of a fresh headless instance.
Share one browser connection across multiple AI clients using HTTP mode.
| unagi-cq/cdp-bridge-mcp | italozucareli/zabbix-observability | lllyasviel/singlefiledb | |
|---|---|---|---|
| Stars | 85 | 85 | 85 |
| Language | Python | Python | Python |
| Last pushed | — | — | 2023-11-24 |
| Maintenance | — | — | Dormant |
| Setup difficulty | moderate | moderate | easy |
| Complexity | 3/5 | 3/5 | 2/5 |
| Audience | developer | ops devops | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires installing a companion Chrome extension alongside the Python server.
CDP Bridge MCP is a Python server that lets an AI coding agent like Claude Code or Codex read and interact with the browser tabs you already have open, using your existing logged-in sessions. The README is written primarily in Chinese. Most browser automation tools like Playwright launch a fresh, separate browser instance with no saved logins or cookies. CDP Bridge MCP works differently: it connects to your real running browser through a companion Chrome extension. Because it reuses your actual session, the AI can immediately access sites you are already logged into without needing to re-authenticate or export credentials. It works using CDP (Chrome DevTools Protocol), the low-level API that Chrome exposes for automation and debugging. The extension installed in your browser connects back to the Python server over a WebSocket connection. When an AI agent calls a tool like browser_scan or browser_execute_js, the server passes the instruction through that WebSocket to the extension, which executes it in the actual page and returns the result. The browser_scan tool strips scripts and styles from the page HTML before returning it, reducing the amount of text sent to the language model. The server implements the MCP (Model Context Protocol), a standard interface for giving AI assistants access to external tools. It supports two connection modes: a simple local mode where the MCP client starts the server as a child process, and an HTTP mode suitable for hosting on a remote server so multiple clients can share one browser connection. A developer building an AI agent that needs to read web content behind login walls, scrape data from authenticated pages, or automate interactions in an existing browser session would use this.
A Python MCP server that lets AI coding agents read and control the browser tabs you already have open, reusing your logged-in sessions.
Mainly Python. The stack also includes Python, Chrome Extension, WebSocket.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.