Back to Blog

Share Your Browser With Cloud AI Agents

One command gives your cloud AI agents secure access to your local Chrome — with your cookies, sessions, and identity.

DV

Dzianis Vashchuk

3 min read

npx @vibebrowser/chrome-devtools-mcp install --port 9333

That one command installs and starts a local MCP bridge so cloud AI agents can use your real Chrome session (tabs, cookies, logins).

Problem

Cloud agents run remotely, but your real browser identity lives on your laptop.

  • Local-only MCP setups break for cloud runtimes
  • Multiple agents can conflict when they all try to share one browser connection
  • You want your existing Chrome session, not a fresh throwaway browser

Solution

Use the forked MCP server from Vibe Browser:

It adds HTTP multi-session support so multiple agents can connect to one Chrome safely. It also supports Tailscale for private cloud-to-local access.

Briefly: this avoids relying on --remote-debugging-port as your primary setup path.

Install

Local-only access:

npx @vibebrowser/chrome-devtools-mcp install --port 9333

Cloud access over Tailscale:

npx @vibebrowser/chrome-devtools-mcp install --port 9333 --tailscale

Configure agents

The installer can auto-configure common clients. Manual examples:

Claude Code

claude mcp add --transport http -s user chrome-devtools http://localhost:9333/mcp

GitHub Copilot CLI (~/.copilot/mcp-config.json)

{
  "mcpServers": {
    "chrome-devtools": {
      "type": "http",
      "url": "http://localhost:9333/mcp"
    }
  }
}

OpenCode (~/.config/opencode/opencode.json)

{
  "mcp": {
    "chrome-devtools": {
      "type": "remote",
      "url": "http://localhost:9333/mcp",
      "enabled": true
    }
  }
}

For cloud agents, replace localhost with your Tailscale hostname.

Done

You now have one local Chrome shared safely across multiple AI agents, with optional private remote access via Tailscale.