flow-mcp

# Flow MCP: Native Google Flow AI Integration This repository contains the configuration and guides to run and connect the Google Flow MCP server remotely. --- ## 📂 Files in this Repository * **`docker-compose.yml`** — Configuration to run the Flow API & MCP server container. * **`mcp`** — Configuration file to connect local Claude Desktop/Cursor to the Docker container. --- ## 🚀 1. Deploying the Server (On Any PC) To start the server: 1. Ensure **Docker** is running. 2. Run this command in the repository folder: ```bash docker-compose up -d ``` 3. Load Chrome and verify that your **Flow Chrome Extension** is connected. *Note: If you deploy on another PC with a different Cloudflare subdomain, update the `PUBLIC_BASE_URL` in `docker-compose.yml` before starting.* --- ## 🔌 2. Connecting to Cursor / Desktop Apps (SSE Transport) To connect your IDE or desktop app (like Cursor) to the server over the internet: Open Cursor Settings (`Cmd + ,`) -> **Models** -> **MCP** -> **Add New MCP Server**: * **Name**: `flow` * **Type**: `SSE` * **URL**: `https://flow.chatbulky.com/sse` --- ## 🌐 3. Connecting to ChatGPT Web (Chrome) For browser-based ChatGPT (Web): 1. Add a Custom Connector/MCP tool in your ChatGPT Web interface. 2. Select **SSE** transport type. 3. Use the URL: ```text https://flow.chatbulky.com/sse ``` 4. Set Authentication to **None** (Disable OAuth). --- ## 🛠️ 4. Connecting Locally via Docker (Claude Desktop / Cursor) If you are running the server locally via Docker, you can connect your Claude Desktop or Cursor client directly to the container using the `mcp` configuration file. Add the following to your `claude_desktop_config.json` (or Cursor MCP config): ```json { "mcpServers": { "flow-agent": { "command": "docker", "args": [ "exec", "-i", "flow-agent-server", "python3", "-u", "/app/flow_mcp_server.py" ] } } } ```