mcp-google-gmail

mcp-google-gmail

A Python-based MCP server that bridges MCP-compatible clients with the Gmail API, enabling AI-driven email management through natural language commands.

Category
访问服务器

README

<div align="center"> <b>mcp-google-gmail</b>

<p align="center"> <i>Your AI Assistant's Gateway to Gmail!</i> 📧 </p>

PyPI - Version PyPI Downloads License: MIT </div>


🤔 What is this?

mcp-google-gmail is a Python-based MCP server that acts as a bridge between any MCP-compatible client (like Claude Desktop, Cursor, or Windsurf) and the Gmail API. It allows you to list, read, search, send, draft, label, and trash emails — all driven by AI through natural language.


🚀 Quick Start

Essentially the server runs in one line: uvx mcp-google-gmail@latest.

This command will automatically download the latest code and run it. We recommend always using @latest to ensure you have the newest version with the latest features and bug fixes.

  1. ☁️ Prerequisite: Google Cloud Setup

  2. 🐍 Install uv

    • uvx is part of uv, a fast Python package installer. Install it if you haven't:
      # macOS / Linux
      curl -LsSf https://astral.sh/uv/install.sh | sh
      # Windows
      powershell -c "irm https://astral.sh/uv/install.ps1 | iex"
      
  3. 🔐 Authenticate

    • Run the built-in auth command to set up your credentials:
      # Point to your OAuth credentials file
      GMAIL_CREDENTIALS_PATH="/path/to/credentials.json" uvx mcp-google-gmail@latest auth
      
    • This opens your browser for Google sign-in. After granting permission, a token.json is saved automatically.
    • You only need to do this once — subsequent runs use the cached token.
  4. 🏃 Run the Server!

    uvx mcp-google-gmail@latest
    
  5. 🔌 Connect your MCP Client

    • Configure your client (e.g., Claude Desktop) to connect to the running server.
    • ➡️ See Usage with Claude Desktop for config examples.

You're ready! Start issuing commands via your MCP client.


✨ Key Features

  • Full Gmail Access: Read, search, send, draft, label, and trash emails.
  • 15 Tools covering all common Gmail operations.
  • Flexible Authentication: Supports OAuth 2.0, Service Accounts, Base64 injection, and Application Default Credentials.
  • Pagination: All list operations support page_token and max_results.
  • Attachments: Send emails with file attachments.
  • Reply Threading: Reply to existing threads with proper In-Reply-To headers.
  • HTML Email: Send plain text and/or HTML bodies.
  • Stdio & SSE Transports: Works with Claude Desktop, Cursor, and remote/container deployments.

🔐 Authentication

The auth Command

Before using the MCP server, authenticate with Gmail:

# Using OAuth credentials (interactive — opens browser)
GMAIL_CREDENTIALS_PATH="/path/to/credentials.json" uvx mcp-google-gmail@latest auth

# Specify where to save the token
GMAIL_CREDENTIALS_PATH="/path/to/credentials.json" \
GMAIL_TOKEN_PATH="/path/to/token.json" \
uvx mcp-google-gmail@latest auth

On success, you'll see:

Credentials path: /path/to/credentials.json
Token path:       token.json

Authenticated as: you@gmail.com
Total messages:   12345
Token saved to:   token.json

Authentication Priority

The server checks for credentials in this order:

  1. GMAIL_CREDENTIALS_CONFIG — Base64-encoded service account JSON (env var)
  2. GMAIL_SERVICE_ACCOUNT_PATH — Path to service account key file
  3. GMAIL_TOKEN_PATH — Path to existing OAuth token
  4. GMAIL_CREDENTIALS_PATH — Path to OAuth credentials (interactive browser flow)
  5. Application Default Credentials — GOOGLE_APPLICATION_CREDENTIALS / gcloud / GCP metadata

Method A: OAuth 2.0 (Personal Accounts) 🧑‍💻

Best for personal use or local development.

  1. Set up OAuth credentials in Google Cloud Console (see GCP Setup)
  2. Run mcp-google-gmail auth to authenticate via browser
  3. Token is cached for future use with automatic refresh
  • GMAIL_CREDENTIALS_PATH — Path to OAuth credentials.json (default: credentials.json)
  • GMAIL_TOKEN_PATH — Where to store the token (default: token.json)

Method B: Service Account (Servers/Automation) ✅

Best for headless environments. Requires domain-wide delegation for accessing user mailboxes.

  • GMAIL_SERVICE_ACCOUNT_PATH — Path to service account JSON key (default: service_account.json)

Method C: Base64-Encoded Credentials (Containers) 🔒

Best for Docker, Kubernetes, CI/CD where file mounting is impractical.

  • GMAIL_CREDENTIALS_CONFIG — Base64-encoded content of your service account JSON
  • Generate with: base64 -w 0 service_account.json

Method D: Application Default Credentials (GCP) 🌐

Best for Google Cloud environments (Cloud Run, GKE, Compute Engine).

  • Uses GOOGLE_APPLICATION_CREDENTIALS or gcloud auth application-default login
  • No additional env vars needed — used as automatic fallback

🛠️ Available Tools (15 Total)

Read Operations

  • gmail_list_messages — List messages with query, labels, pagination (1-500 per page)

    • query (optional): Gmail search query (e.g. "is:unread", "from:alice@example.com")
    • label_ids (optional): Filter by label IDs (e.g. ["INBOX"])
    • max_results (optional, default 20): Messages per page (1-500)
    • page_token (optional): Token for next page
    • include_spam_trash (optional, default false): Include spam/trash
    • Returns: {messages: [{id, thread_id, snippet, subject, from, date}], next_page_token, result_size_estimate}
  • gmail_get_message — Get full message by ID (headers, body, attachments)

    • message_id: The Gmail message ID
    • Returns: {id, thread_id, subject, from, to, cc, date, body_text, body_html, labels, attachments}
  • gmail_search_messages — Search with Gmail query syntax, returns compact summaries (1-100 per page). Use gmail_get_message to read the full body of a result.

    • query: Gmail search query (e.g. "has:attachment after:2024/01/01")
    • max_results (optional, default 10): Results per page (1-100)
    • page_token (optional): Token for next page
    • Returns: {messages: [{id, thread_id, snippet, subject, from, to, date, labels}], next_page_token, result_size_estimate}
  • gmail_list_drafts — List drafts with pagination and query filter

    • max_results (optional, default 20): Drafts per page (1-500)
    • page_token (optional): Token for next page
    • query (optional): Gmail search query to filter drafts
    • Returns: {drafts: [{draft_id, message_id, subject, to, snippet}], next_page_token, result_size_estimate}

Send Operations

  • gmail_send_message — Send email with to/cc/bcc, HTML body, attachments, reply threading

    • to, subject, body (required)
    • cc, bcc, html_body, attachment_paths (optional)
    • reply_to_message_id, thread_id (optional, for threading)
    • Returns: {id, thread_id, label_ids}
  • gmail_create_draft — Create a draft without sending (same params as send)

    • Returns: {draft_id, message_id}
  • gmail_update_draft — Update an existing draft (merges provided fields with existing)

    • draft_id (required), all other fields optional
    • Returns: {draft_id, message_id}
  • gmail_delete_draft — Permanently delete a draft

    • draft_id (required)
    • Returns: {success: true}
  • gmail_send_draft — Send an existing draft

    • draft_id (required)
    • Returns: {message_id, thread_id}

Label Operations

  • gmail_list_labels — List all labels (system and user-created)

    • Returns: {labels: [{id, name, type}]}
  • gmail_create_label — Create a new label (supports nesting with /)

    • name: Label name (e.g. "Projects/Work")
    • Returns: {id, name}
  • gmail_delete_label — Delete a user label (system labels cannot be deleted)

    • label_id: The label ID
    • Returns: {success: true}
  • gmail_modify_message_labels — Add/remove labels from a message

    • message_id (required)
    • add_label_ids (optional): Label IDs to add
    • remove_label_ids (optional): Label IDs to remove
    • Returns: {id, label_ids}

Trash Operations

  • gmail_trash_message — Move a message to trash (auto-deleted after 30 days)

    • message_id: The message ID
    • Returns: {id, label_ids}
  • gmail_untrash_message — Restore a message from trash

    • message_id: The message ID
    • Returns: {id, label_ids}

🔌 Usage with Claude Desktop

Add the server config to your claude_desktop_config.json:

<details> <summary>🔵 Config: uvx + OAuth (Recommended for personal use)</summary>

{
  "mcpServers": {
    "gmail": {
      "command": "uvx",
      "args": ["mcp-google-gmail@latest"],
      "env": {
        "GMAIL_CREDENTIALS_PATH": "/path/to/credentials.json",
        "GMAIL_TOKEN_PATH": "/path/to/token.json"
      }
    }
  }
}

🍎 macOS Note: If you get a spawn uvx ENOENT error, use the full path:

"command": "/Users/yourusername/.local/bin/uvx"

</details>

<details> <summary>🔵 Config: uvx + Service Account</summary>

{
  "mcpServers": {
    "gmail": {
      "command": "uvx",
      "args": ["mcp-google-gmail@latest"],
      "env": {
        "GMAIL_SERVICE_ACCOUNT_PATH": "/path/to/service_account.json"
      }
    }
  }
}

</details>

<details> <summary>🟡 Config: Development (from cloned repo)</summary>

{
  "mcpServers": {
    "gmail": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/mcp-gmail", "mcp-google-gmail"]
    }
  }
}

</details>


⚙️ Usage with Cursor / Windsurf

{
  "mcpServers": {
    "gmail": {
      "command": "uvx",
      "args": ["mcp-google-gmail@latest"],
      "env": {
        "GMAIL_CREDENTIALS_PATH": "/path/to/credentials.json",
        "GMAIL_TOKEN_PATH": "/path/to/token.json"
      }
    }
  }
}

🐳 SSE Transport (Remote / Container)

uv run mcp-google-gmail --transport sse
Variable Default Description
HOST / FASTMCP_HOST 0.0.0.0 Bind address
PORT / FASTMCP_PORT 8000 Listen port

☁️ Google Cloud Platform Setup (Detailed)

This setup is required before running the server.

  1. Create/Select a GCP Project — Go to the Google Cloud Console
  2. Enable the Gmail API — Navigate to "APIs & Services" → "Library", search for "Gmail API", click Enable
  3. Configure OAuth Consent Screen — Go to "APIs & Services" → "OAuth consent screen", select External, fill in app name and contact email, add the scope https://www.googleapis.com/auth/gmail.modify
  4. Create OAuth Credentials — Go to "APIs & Services" → "Credentials" → "Create Credentials" → "OAuth 2.0 Client ID", select Desktop application
  5. Download Credentials — Click the download button and save as credentials.json
  6. Authenticate — Run the auth command:
    GMAIL_CREDENTIALS_PATH="/path/to/credentials.json" uvx mcp-google-gmail@latest auth
    
    Complete the browser sign-in. A token.json will be saved for future use.

For Service Accounts: Go to Credentials → Create Credentials → Service Account, create a key (JSON), download it. Note: Service accounts require domain-wide delegation for Gmail access.


🔧 Environment Variables Reference

Variable Default Description
GMAIL_CREDENTIALS_CONFIG — Base64-encoded service account JSON
GMAIL_SERVICE_ACCOUNT_PATH service_account.json Path to service account key file
GMAIL_TOKEN_PATH token.json Path to OAuth token file
GMAIL_CREDENTIALS_PATH credentials.json Path to OAuth client credentials file
HOST / FASTMCP_HOST 0.0.0.0 SSE transport bind address
PORT / FASTMCP_PORT 8000 SSE transport port

💬 Example Prompts for Claude

  • "List my 10 most recent unread emails"
  • "Search for emails from alice@example.com with attachments"
  • "Send an email to bob@example.com with subject 'Meeting Notes' and the body 'Here are the notes from today.'"
  • "Create a draft reply to the last email from my manager"
  • "Label all emails from newsletter@example.com as 'Newsletters'"
  • "Trash all promotional emails from the last week"
  • "Show me the full content of message ID abc123"
  • "What are my unread emails about project deadlines?"

🤝 Contributing

Contributions are welcome! Please open an issue to discuss bugs or feature requests. Pull requests are appreciated.


📄 License

This project is licensed under the MIT License — see the LICENSE file for details.


🙏 Credits

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