VibeBrowser Co-Pilot for Multi-Agent Builders

AI for Claude + Codex

Chrome exposes exactly one DevTools Protocol (CDP) debugger attachment per tab. When a second AI agent tries to attach, Chrome silently drops the first — and you get "Target closed." VibeBrowser Co-Pilot is a Chrome extension that holds the single chrome.debugger session and fans out commands to every connected MCP client. Drop it in; the conflicts disappear.

Get Started in 30 Seconds

One command wires up a cloud browser to your AI tool — then hand it any browser task.

multi-agent-setup
$
Snapshot output is markdown — compact, structured, perfect for agent context windows

Why Multi-Agent Builders Choose Vibe

CDP multiplexer, inside Chrome

One stable chrome.debugger session shared across all agents. Each MCP client connects to the extension — no port management, no external proxy.

Drop-in for any MCP client

Works with Claude Desktop, Claude Code, Codex CLI, custom agents, and any WebSocket-based automation client. No code changes to your agents.

Your real browser, your sessions

Agents run inside your actual Chrome profile. Cookies, logins, and auth state stay intact across all concurrent agent sessions.

Zero sidecar processes

Runs entirely inside Chrome via the extensions API. No extra Node server, no CDP proxy daemon, no additional ports to manage.

Key Capabilities

CDP multiplexer, inside Chrome

One stable chrome.debugger session shared across all agents. Each MCP client connects to the extension — no port management, no external proxy.

Drop-in for any MCP client

Works with Claude Desktop, Claude Code, Codex CLI, custom agents, and any WebSocket-based automation client. No code changes to your agents.

Your real browser, your sessions

Agents run inside your actual Chrome profile. Cookies, logins, and auth state stay intact across all concurrent agent sessions.

Zero sidecar processes

Runs entirely inside Chrome via the extensions API. No extra Node server, no CDP proxy daemon, no additional ports to manage.

Survives agent restarts

If one agent disconnects or crashes, the shared session stays open. Other agents keep working without re-attaching to Chrome.

Watch all agents live

Extension popup shows every connected client and active commands in real time. Debug multi-agent workflows without guessing which agent did what.

Common Workflows

Run Claude Code + a custom MCP scraper in the same Chrome tab without either dropping

Use a planner agent and an executor agent simultaneously on the same authenticated session

Attach Playwright tests alongside an AI agent without "browser already running" errors

Run a monitoring agent and an action agent in the same browser 24/7

Let multiple AI coding assistants share the same browser for E2E testing

Connect any combination of MCP clients to one real, logged-in Chrome session

Integrations & Agent Ecosystem

Google Workspace, MCP access for other agents, reusable skills, and a secure secrets vault.

Google Workspace Native

Gmail + Calendar actions built in for real workflows, not brittle APIs.

MCP Server for Agents

Expose your browser as an MCP server so other AI agents can drive it.

Remote Browser Relay

Securely connect remote agents to your logged-in browser without VPNs.

Skills Library

Create reusable skills and share them across your team or clients.

Secrets Vault + Type-In

Internal vault with a password-fill tool that never exposes secrets to the LLM.

Model & Agent Choice

Works with Vibe AI, Anthropic Claude Max, GitHub Copilot, and BYOK providers.

Works With Your Tools

Claude Desktop
Claude Code
Codex CLI
Cursor
Windsurf
Custom MCP servers
Playwright MCP
Chrome DevTools MCP
Any WebSocket client

Frequently Asked Questions

Why does Chrome drop the first agent when a second one connects?

Chrome's DevTools Protocol only allows one debugger attachment per tab at a time. When a second tool calls Target.attachToTarget (or any equivalent), Chrome silently revokes the first attachment — causing "Target closed" or "Session invalidated" errors in whatever was connected first.

How does VibeBrowser Co-Pilot fix this?

The extension claims the single chrome.debugger slot itself at startup. All MCP clients then connect to the extension's relay port rather than to Chrome directly. The extension proxies commands and events between all clients and Chrome — so every agent thinks it has its own connection, and none of them ever see a conflict.

Will this work with Claude Desktop + Claude Code running at the same time?

Yes. Both use MCP. Point both to the Co-Pilot relay endpoint. They'll share the same Chrome session without interfering with each other.

Does it work with Playwright MCP and Chrome DevTools MCP simultaneously?

Yes — as long as both connect via the extension relay. If Playwright MCP tries to launch its own Chrome instance, that's a separate process; to share one browser, configure Playwright MCP to connect to an existing Chrome via the extension.

Is this a cloud product or local?

Local-first. The extension runs inside your Chrome on your machine. There's no data leaving your browser unless you opt into the VibeBrowser Cloud relay for remote access.

Does the Chrome Web Store block sideloading?

No. Chrome supports loading unpacked extensions (Developer Mode) and .crx installs, which is how you install Co-Pilot while the Web Store review is in progress. Takes about 2 minutes.