Google Docs/Sheets MCP Server
A remote MCP server that lets Claude edit existing Google Docs and Sheets, supporting search, replace, append, and insert operations for Docs, and search, update, and append rows for Sheets.
README
Google Docs/Sheets MCP Server
A remote MCP server that lets Claude actually edit your existing Google Docs and Sheets — not just read them or create new ones.
Tools it exposes:
docs_search,docs_read,docs_replace_text,docs_append_text,docs_insert_text_at,docs_createsheets_search,sheets_read,sheets_update_cells,sheets_append_rows,sheets_create
This server acts as your Google account only (whichever account you authorize in Step 2). It's meant for personal or single-team use, not as a multi-tenant public service.
Step 1 — Create a Google Cloud OAuth client
- Go to https://console.cloud.google.com/ and create a project (or use an existing one).
- Go to APIs & Services → Library and enable:
- Google Docs API
- Google Sheets API
- Google Drive API
- Go to APIs & Services → OAuth consent screen. Choose External, fill in the required fields, and add yourself as a test user. (You don't need to publish the app — test mode works fine for personal use.)
- Go to APIs & Services → Credentials → Create Credentials → OAuth client ID.
- Application type: Desktop app
- Save the generated Client ID and Client Secret.
Step 2 — Get a refresh token (run this on your own computer)
cd gdocs-mcp-server
npm install
GOOGLE_CLIENT_ID=your_client_id \
GOOGLE_CLIENT_SECRET=your_client_secret \
npm run get-token
This opens a URL for you to visit — log in and approve access. The script
prints a GOOGLE_REFRESH_TOKEN value in your terminal. Save it; you'll need
it in Step 3.
Step 3 — Deploy the server
Pick any host that runs a persistent Node.js process and gives you a public HTTPS URL. Render.com is a straightforward free option:
- Push this folder to a GitHub repo (or use Render's "public Git repo" deploy).
- On Render: New → Web Service, connect the repo.
- Build command:
npm install - Start command:
npm start - Add environment variables (from Environment tab):
GOOGLE_CLIENT_IDGOOGLE_CLIENT_SECRETGOOGLE_REFRESH_TOKEN(from Step 2)MCP_ACCESS_TOKEN— make up a long random secret (e.g. runopenssl rand -hex 32locally)
- Deploy. Render will give you a URL like
https://your-app.onrender.com.
Your MCP endpoint is then: https://your-app.onrender.com/mcp
(Railway, Fly.io, or a small VPS all work the same way — the only requirement is a public HTTPS URL and the same environment variables.)
Step 4 — Connect it to Claude
- In Claude, go to Settings → Connectors.
- Click "+" → Add custom connector.
- Paste in your server URL:
https://your-app.onrender.com/mcp - Click Advanced settings → under Request headers, add:
- Header name:
Authorization - Header value:
Bearer YOUR_MCP_ACCESS_TOKEN(the same secret from Step 3)
- Header name:
- Click Add, then Connect.
- In any chat, use the "+" button → Connectors → toggle this one on.
You can now ask Claude things like "update cell B2 in my Q3 budget sheet to 5000" or "replace 'draft' with 'final' throughout my proposal doc" and it will actually make the change.
Testing locally before you deploy
cp .env.example .env
# fill in .env with your values
npm install
npm start
The server listens on http://localhost:3000/mcp. You won't be able to point
Claude's hosted product at localhost directly (it needs a public URL), but
this is useful for catching errors before you deploy.
Notes and limitations
- This server edits Docs/Sheets as whichever Google account you authorized in Step 2 — anyone using this connector acts as that account.
docs_replace_textreplaces plain text only; it won't preserve rich formatting differences between the old and new text.- Sheets values are written with
USER_ENTEREDinput, so formulas like=SUM(A1:A10)are interpreted, not stored as literal text. - If you ever want to revoke access, go to https://myaccount.google.com/permissions and remove the app.
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。