Connect cdpfleet to your AI assistant

cdpfleet's MCP server gives your assistant a real remote browser — the browser and proxy of your choice, a live view a person can take over, a recording of what happened. Pick your assistant below; then paste the prompt from the code generator's AI · MCP tab.

The short version: add https://cdpfleet.com/mcp as an MCP server or connector, and when your assistant opens a cdpfleet window, sign in and click Allow (OAuth). Nothing to copy, and you can disconnect it any time in dashboard → API keys → Connected AI apps.

Scripts and assistants without OAuth can send an API key in an Authorization: Bearer header — keep it private.

Desktop and web apps

Claude — claude.ai and Claude Desktop

1. Settings → Connectors → Add custom connector.
2. Name: cdpfleet. Remote MCP server URL: https://cdpfleet.com/mcp. Add, then Connect: sign in to cdpfleet and click Allow.
3. In a chat, turn the cdpfleet connector on (the tools menu under the message box) and paste your prompt.

Custom connectors depend on your Claude plan; on Team and Enterprise an owner may need to allow them.

ChatGPT — chatgpt.com and the desktop app

1. Settings → Apps & Connectors → Advanced settings → turn on Developer mode.
2. Back in Apps & Connectors → Create. Name: cdpfleet. MCP server URL: https://cdpfleet.com/mcp. Authentication: OAuth. Create, then sign in to cdpfleet and click Allow.
3. In a chat, add the cdpfleet connector (the + menu → Developer mode) and paste your prompt.

Developer mode and custom connectors depend on your ChatGPT plan and workspace settings.

Gemini — gemini.google.com

1. Add a custom app (MCP server) with the URL https://cdpfleet.com/mcp.
2. When Gemini asks you to link your account, sign in to cdpfleet and click Allow.
3. In a chat, use the cdpfleet app and paste your prompt.

Coding assistants

Claude Code

claude mcp add --transport http cdpfleet https://cdpfleet.com/mcp

Then run /mcp inside Claude Code and choose Authenticate: a browser window opens to sign in and allow. Or skip OAuth with an API key: add --header "Authorization: Bearer YOUR_API_KEY".

Codex (OpenAI's CLI)

Add to ~/.codex/config.toml — this runs the same server locally with npx (Node.js 18+):

[mcp_servers.cdpfleet]
command = "npx"
args = ["-y", "cdpfleet-mcp"]
env = { CDPFLEET_API_KEY = "YOUR_API_KEY" }

Cursor

Settings → MCP → Add new MCP server, or ~/.cursor/mcp.json — Cursor signs you in with OAuth on first use:

{
  "mcpServers": {
    "cdpfleet": {
      "url": "https://cdpfleet.com/mcp"
    }
  }
}

Or add "headers": { "Authorization": "Bearer YOUR_API_KEY" } to skip OAuth.

Other AI assistants

VS Code (GitHub Copilot), Windsurf, Gemini CLI, Cline and most other MCP clients take one of these:

• Remote server (Streamable HTTP): URL https://cdpfleet.com/mcp — with OAuth if the client supports it, otherwise the header Authorization: Bearer YOUR_API_KEY.
• Local server: command npx -y cdpfleet-mcp with the environment variable CDPFLEET_API_KEY.

Then

Open the code generator, choose the browser, proxy and options, switch to AI · MCP, type what you want done and copy the prompt into your assistant. The tools it uses — launch_browser, navigate, snapshot, click, type, screenshot, live_view, close_browser — are listed on the MCP server page. Browsers are billed and limited by your plan as from code.