Google Flow MCP
Bridges AI agents like Codex and Claude to Google Flow via browser automation. Provides tools to open, snapshot, click, type, upload, download, wait, and confirm paid generations.
README
Google Flow MCP
Clean MCP server for Google Flow (labs.google/fx/tools/flow) — a thin browser bridge that lets AI agents (Codex, Claude Code, Antigravity) control Google Flow directly using your Google Pro account.
No API. No third-party services. Just your browser, your account, your credits.
What It Can Do
| Tool | What it does |
|---|---|
flow_open |
Navigate to Flow, or a specific project |
flow_snapshot |
See all UI elements and generated media on screen |
flow_click |
Click any button, menu item, or card |
flow_type |
Type a prompt into any input field |
flow_upload |
Upload a local file (reference image, etc.) |
flow_download |
Save a generated image or video to disk |
flow_wait |
Wait for generation to finish |
flow_confirm_paid_generation |
Authorize a paid Veo/Omni generation (safety guard) |
Setup
1. Install & Build
cd "/Volumes/Xstorage/MCP - Googel Flow 2"
npm install
npm run build
2. Configure Environment
cp .env.example .env
Edit .env — set LOCAL_STORAGE_ROOT to wherever you want images saved.
3. One-Time Google Login
npm run login
Chrome opens. Sign into your Google account and navigate to labs.google/fx/tools/flow.
Close the script with Ctrl+C when done. Your session persists forever (or until Google forces re-auth).
Platform Integration
Antigravity IDE
Add to ~/.gemini/antigravity-ide/mcp_config.json:
{
"mcpServers": {
"google-flow": {
"command": "node",
"args": ["/Volumes/Xstorage/MCP - Googel Flow 2/dist/index.js"],
"env": {
"LOCAL_STORAGE_ROOT": "/Volumes/Xstorage/Media",
"HEADLESS": "false"
}
}
}
}
Codex (ChatGPT Desktop)
Add to ~/.codex/config.toml:
[mcp_servers.google-flow]
command = "node"
args = ["/Volumes/Xstorage/MCP - Googel Flow 2/dist/index.js"]
startup_timeout_sec = 30
[mcp_servers.google-flow.env]
LOCAL_STORAGE_ROOT = "/Volumes/Xstorage/Media"
HEADLESS = "false"
Restart Codex to pick up.
Claude Code (Desktop App)
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"google-flow": {
"command": "node",
"args": ["/Volumes/Xstorage/MCP - Googel Flow 2/dist/index.js"],
"env": {
"LOCAL_STORAGE_ROOT": "/Volumes/Xstorage/Media",
"HEADLESS": "false"
}
}
}
}
Restart Claude to pick up.
Example Agent Workflow
Tell the agent (Codex, Claude, Antigravity):
"Open Google Flow, type this prompt into the image composer: 'cinematic wide shot of ancient Greece at sunset, photorealistic', generate the image, wait for it to finish, and download it to /Volumes/Xstorage/Media"
The agent will chain:
flow_open → flow_snapshot → flow_type → flow_click → flow_wait → flow_download
For paid Veo video generation, the agent will call flow_confirm_paid_generation first and ask your approval.
Development
npm run dev # Run without building (tsx)
npm run build # Compile to dist/
npm run typecheck # Type check only
npm run login # One-time Chrome Google login
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。