Gmail Server for Model Context Protocol (MCP)
This MCP server integrates with Gmail to enable sending, removing, reading, drafting, and responding to emails.
Note: This server enables an MCP client to read, remove, and send emails. However, the client prompts the user before conducting such activities.
https://github.com/user-attachments/assets/5794cd16-00d2-45a2-884a-8ba0c3a90c90
Components
Tools
-
send-email
- Sends email to email address recipient
- Input:
recipient_id(string): Email address of addresseesubject(string): Email subjectmessage(string): Email content
- Returns status and message_id
-
trash-email
- Moves email to trash
- Input:
email_id(string): Auto-generated ID of email
- Returns success message
-
mark-email-as-read
- Marks email as read
- Input:
email_id(string): Auto-generated ID of email
- Returns success message
-
get-unread-emails
- Retrieves unread emails
- Returns list of emails including email ID
-
read-email
- Retrieves given email content
- Input:
email_id(string): Auto-generated ID of email
- Returns dictionary of email metadata and marks email as read
-
open-email
- Open email in browser
- Input:
email_id(string): Auto-generated ID of email
- Returns success message and opens given email in default browser
Setup
Gmail API Setup
- Create a new Google Cloud project
- Enable the Gmail API
- Configure an OAuth consent screen
- Select "external". However, we will not publish the app.
- Add your personal email address as a "Test user".
- Add OAuth scope
https://www.googleapis.com/auth/gmail/modify - Create an OAuth Client ID for application type "Desktop App"
- Download the JSON file of your client's OAuth keys
- Rename the key file and save it to your local machine in a secure location. Take note of the location.
- The absolute path to this file will be passed as parameter
--creds-file-pathwhen the server is started.
- The absolute path to this file will be passed as parameter
Authentication
When the server is started, an authentication flow will be launched in your system browser.
Token credentials will be subsequently saved (and later retrieved) in the absolute file path passed to parameter --token-path.
For example, you may use a dot directory in your home folder, replacing [your-home-folder].:
| Parameter | Example |
|---|---|
--creds-file-path |
/[your-home-folder]/.google/client_creds.json |
--token-path |
/[your-home-folder]/.google/app_tokens.json |
Usage with Desktop App
Using uv is recommended.
To integrate this server with Claude Desktop as the MCP Client, add the following to your app's server configuration. By default, this is stored as ~/Library/Application\\ Support/Claude/claude_desktop_config.json.
{
"mcpServers": {
"gdrive": {
"command": "uv",
"args": [
"--directory",
"[absolute-path-to-git-repo]",
"run",
"gmail",
"--creds-file-path",
"[absolute-path-to-credentials-file]",
"--token-path",
"[absolute-path-to-access-tokens-file]"
]
}
}
}
The following parameters must be set
| Parameter | Example |
|---|---|
--directory |
Absolute path to gmail directory containing server |
--creds-file-path |
Absolute path to credentials file created in Gmail API Setup. |
--token-path |
Absolute path to store and retrieve access and refresh tokens for application. |
Troubleshooting with MCP Inspector
To test the server, use MCP Inspector. From the git repo, run the below changing the parameter arguments accordingly.
npx @modelcontextprotocol/inspector uv run [absolute-path-to-git-repo]/src/gmail/server.py --creds-file-path [absolute-path-to-credentials-file] --token-path [absolute-path-to-access-tokens-file]
Recommend MCP Servers 💡
Storyblok MCP
An MCP implementation for Storyblok that enables managing components using natural language descriptions.
nx-mcp
The Nx MCP server enhances LLMs by providing rich workspace metadata, enabling them to understand monorepo architecture and make context-aware decisions for improved AI-powered development.
filesystem-mcp-server
Empower your AI agents with robust, platform-agnostic file system capabilities, now with STDIO & Streamable HTTP transport options.
root-signals
An MCP server that exposes Root Signals evaluators as tools, enabling AI assistants and agents to evaluate LLM responses against various quality criteria.
g-search-mcp
A powerful MCP server for Google search that enables parallel searching with multiple keywords simultaneously.
@zcaceres/gtasks
A Google Tasks Model Context Protocol Server for Claude