mcp-gdrive-cf

mcp-gdrive-cf

Remote MCP server for Google Drive and Sheets running on Cloudflare Workers, providing 11 tools for file and sheet operations with full OAuth 2.0 and PKCE support for secure authentication.

Category
访问服务器

README

mcp-gdrive-cf

Remote MCP server for Google Drive and Sheets running on Cloudflare Workers with full OAuth 2.0 support.

Production Ready ✅ | 11 Tools | OAuth 2.0 | PKCE | Global Edge Network

Adapted from isaacphi/mcp-gdrive to run as a remote HTTP/SSE MCP server on Cloudflare's edge network with complete OAuth 2.0 authorization server capabilities.


🚀 Quick Start

```bash

1. Clone and install

git clone https://github.com/brianmoney/mcp-gdrive-cf.git cd mcp-gdrive-cf npm install

2. Create KV namespaces

wrangler kv:namespace create KV_TOKENS wrangler kv:namespace create KV_CLIENTS

3. Update wrangler.toml with your namespace IDs

4. Configure Google OAuth credentials

wrangler secret put GOOGLE_CLIENT_SECRET

5. Deploy

wrangler deploy

6. Test with MCP Inspector

npx @modelcontextprotocol/inspector https://your-worker.workers.dev/sse ```


✨ Features

Drive Operations

  • 🔍 Search - Basic and advanced search with filters (MIME type, owner, dates, shared drives)
  • 📄 Read - Read any file with 22+ export formats (PDF, DOCX, XLSX, Markdown, etc.)
  • 📁 Create Folders - Organize files with nested folder structures
  • ⬆️ Upload - Upload files up to 5MB
  • 🗑️ Delete - Move files to trash
  • 📦 Move - Reorganize files between folders
  • 🔐 Share - Add permissions (reader, writer, commenter, owner)

Sheets Operations

  • 📊 Read - Batch read multiple ranges
  • ✏️ Update - Update individual cells
  • Append - Add rows to spreadsheets

Infrastructure

  • ☁️ Global Edge Network - Runs on Cloudflare Workers worldwide
  • 🔐 OAuth 2.0 Server - Full authorization server with PKCE support
  • 🔑 Dynamic Client Registration - Automatic client onboarding
  • 🔄 Auto Token Refresh - Seamless Google token renewal
  • 📡 SSE Transport - Real-time Server-Sent Events
  • 🛡️ Secure - HTTPS only, encrypted token storage

🎯 What Makes This Different

Unlike the original STDIO-based `mcp-gdrive`, this implementation:

  1. Runs remotely on Cloudflare Workers (no local process)
  2. Full OAuth 2.0 authorization server (not just OAuth client)
  3. Multi-client support via dynamic client registration
  4. PKCE security for public clients
  5. 22+ export formats for Drive files
  6. Write operations (create, delete, move, share, upload)
  7. Production tested with real workloads

📋 Prerequisites

Required

  • Cloudflare Account with Workers enabled (free tier works)
  • Google Cloud Project with Drive API and Sheets API enabled
  • Node.js LTS (v18+)
  • Wrangler CLI (`npm install -g wrangler`)

Google Cloud Setup

  1. Create a project at Google Cloud Console
  2. Enable APIs:
    • Google Drive API
    • Google Sheets API
  3. Create OAuth 2.0 credentials:
    • Type: Web application
    • Redirect URIs: `https://your-worker.workers.dev/google/callback`
  4. Required OAuth scopes:
    • `https://www.googleapis.com/auth/drive`
    • `https://www.googleapis.com/auth/spreadsheets`

🔧 Installation & Deployment

1. Install Dependencies

```bash npm install ```

2. Create KV Namespaces

```bash

For user session tokens

wrangler kv:namespace create KV_TOKENS

For OAuth client registrations

wrangler kv:namespace create KV_CLIENTS ```

Copy the namespace IDs to `wrangler.toml`:

```toml [[kv_namespaces]] binding = "KV_TOKENS" id = "your-tokens-namespace-id"

[[kv_namespaces]] binding = "KV_CLIENTS" id = "your-clients-namespace-id" ```

3. Configure Environment Variables

Add to `wrangler.toml`:

```toml [vars] GOOGLE_CLIENT_ID = "your-client-id.apps.googleusercontent.com" ```

Set secrets:

```bash

Google OAuth client secret

wrangler secret put GOOGLE_CLIENT_SECRET ```

4. Deploy

```bash wrangler deploy ```

Your worker will be available at: `https://your-worker-name.your-account.workers.dev`


🔐 OAuth 2.0 Flow

This server implements a complete OAuth 2.0 authorization server per the MCP specification.

Using MCP Inspector

The easiest way to test the OAuth flow:

```bash npx @modelcontextprotocol/inspector https://your-worker.workers.dev/sse ```

  1. Discovery - Inspector auto-discovers OAuth endpoints
  2. Registration - Click "Guided Setup" → registers client automatically
  3. Authorization - Redirects to Google → authenticates → redirects back
  4. Token Exchange - Automatically exchanges code for access token
  5. Use Tools - All 11 tools now available!

Security Features

  • PKCE (RFC 7636) - Proof Key for Code Exchange
  • State Parameter - CSRF protection
  • HTTPS Only - Enforced by Cloudflare Workers
  • Secure Token Storage - Encrypted in Workers KV
  • Token Expiration - Access tokens expire in 1 hour
  • Client Secret Hashing - SHA-256 hashed storage
  • Authorization Code Single-Use - Codes deleted after exchange

🛠️ Available Tools

Drive Tools

  • `gdrive_search` - Basic search across all files
  • `gdrive_search_advanced` - Advanced search with filters
  • `gdrive_read_file` - Read files with 22+ export formats
  • `gdrive_create_folder` - Create folders
  • `gdrive_upload_file` - Upload files (up to 5MB)
  • `gdrive_delete_file` - Move files to trash
  • `gdrive_move_file` - Move files between folders
  • `gdrive_add_permission` - Share files/folders

Sheets Tools

  • `gsheets_read` - Read multiple ranges
  • `gsheets_update_cell` - Update single cells
  • `gsheets_append_row` - Append rows

🔌 Client Configuration

VS Code with MCP Extension

Create `.vscode/settings.json`:

```json { "mcp.servers": { "gdrive": { "url": "https://your-worker.workers.dev/sse", "authorization": { "type": "oauth2" } } } } ```

Claude Desktop / Cline

macOS/Linux: `~/Library/Application Support/Claude/claude_desktop_config.json`
Windows: `%APPDATA%\Claude\claude_desktop_config.json`

```json { "mcpServers": { "gdrive": { "url": "https://your-worker.workers.dev/sse" } } } ```


🧪 Testing

Local Development

```bash npm start

Server available at http://localhost:8788

```

Test with MCP Inspector

```bash npx @modelcontextprotocol/inspector http://localhost:8788/sse ```

Check CloudFlare Logs

```bash wrangler tail --format pretty ```


🐛 Troubleshooting

"Failed to discover OAuth metadata"

Solution:

  • Ensure `/.well-known/oauth-authorization-server` returns valid JSON
  • Check CORS headers are present
  • Verify latest version is deployed: `wrangler deploy`

"Token exchange failed: HTTP 400"

Solutions:

  • Check code_verifier matches original code_challenge
  • Ensure redirect_uri exactly matches registration
  • Verify client credentials are correct
  • Check CloudFlare logs: `wrangler tail`

Tools not working

Solutions:

  • Check Google OAuth scopes include `drive` and `spreadsheets`
  • Re-authenticate if scopes changed: visit `/google/authorize`
  • Verify Worker has valid Google access token
  • Check KV namespace bindings in `wrangler.toml`

📊 Performance & Limits

Resource Limit Notes
Request Timeout 30 seconds CloudFlare Workers limit
File Upload 5 MB Simple upload API limit
File Export 10 MB Google Drive export limit
Token Storage 30 days Automatic cleanup
Authorization Code 10 minutes Single-use
Access Token 1 hour Auto-refreshed

🔒 Security Best Practices

  1. Never commit secrets - Use `wrangler secret put`
  2. Rotate client secrets - Periodically regenerate OAuth credentials
  3. Monitor access - Use CloudFlare Analytics to track usage
  4. Limit OAuth scopes - Only request necessary Google permissions
  5. Use PKCE - Always use PKCE for public clients
  6. Validate redirect URIs - Whitelist exact URIs in Google Console
  7. Enable 2FA - Protect your CloudFlare and Google accounts

📄 License

MIT License


🙏 Acknowledgments


Made with ☁️ by Brian Money

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选