Gmail MCP

Gmail MCP

Local MCP server providing Gmail tools (search, read, draft, send, trash) via Google's official API, integrated with GitHub Copilot CLI.

Category
Visit Server

README

Gmail MCP

CI

Local Gmail tools for GitHub Copilot CLI, backed by Google's official Gmail API.

This project is a standalone local MCP server, not a Copilot plugin. Copilot starts it as a user-configured stdio process.

[!IMPORTANT] OAuth credentials and direct Gmail API traffic remain on this PC, but email content returned by MCP tools may be sent to GitHub Copilot's hosted model services. Do not use this integration if mailbox content must never leave the computer.

Documentation

Safety model

  • Requests gmail.modify for mailbox actions and gmail.settings.basic for filter management, never the permanent-delete https://mail.google.com/ scope.
  • Sending is draft-first. gmail_send_draft returns a state-bound confirmation token before it can send.
  • Deletion accepts message IDs only and moves them to recoverable Gmail Trash.
  • OAuth refresh tokens are stored in Windows Credential Manager. Access tokens remain in memory.
  • Write tools are marked as mutating/destructive for Copilot confirmations. Keep confirmations enabled and do not use /allow-all with this server.
  • Persistent filters cannot forward mail or automate Trash/Spam actions.
  • Attachment download is explicit, bounded to 25 MiB, writes only to an existing absolute local directory, and never overwrites a file.

Google Cloud setup

See Setup and operations for the complete Google Cloud workflow, including multiple-account selection, test users, scope setup, missing JSON downloads, client-secret compatibility, publishing, and cleanup.

Quick start

git clone https://github.com/Evyatar108/gmail-mcp.git
Set-Location gmail-mcp
.\scripts\setup-copilot.ps1
.\scripts\install-oauth-client.ps1 -Path C:\Downloads\client_secret.json
.\.venv\Scripts\gmail-mcp.exe auth
.\.venv\Scripts\gmail-mcp.exe status

Authorization opens the system browser, requests offline access, verifies the authenticated account and granted scopes, and stores the refresh token in Windows Credential Manager.

setup-copilot.ps1 registers the exact 16-tool allowlist using the cloned repository's absolute executable path and installs the three public safety/setup skills.

The server is non-interactive. If authorization is absent or revoked, Gmail tools return an error directing you to run gmail-mcp auth; the stdio server never launches a browser.

Typical flow

  1. Search: Use gmail_search to find receipts from last month.
  2. Read: Use gmail_get_message for message ID ...
  3. Download: select attachment metadata, then save that exact attachment to an existing absolute directory with gmail_download_attachment.
  4. Draft: Create a draft reply, but do not send it.
  5. Send: call gmail_send_draft without confirmation to receive the exact preview token, review it, then call again with that token.
  6. Trash: call gmail_trash without confirmation to receive the exact preview token, review the listed message IDs, then call again with that token.
  7. Rules: create a label, preview a safe filter, then approve the native destructive-tool prompt before filter creation or deletion.

Gmail Trash retention is time-limited. gmail_untrash works only while Gmail still retains the message.

For complete schemas, limits, and confirmation behavior, see Tool reference.

Credential status and revocation

.\.venv\Scripts\gmail-mcp.exe status
.\.venv\Scripts\gmail-mcp.exe revoke

Revocation removes the Google grant and then deletes the WinVault entry. Already in-flight Google requests cannot be cancelled, and an issued access token may remain accepted briefly after revocation. Each new tool call rechecks WinVault and fails closed once the record is removed or changed.

Development

uv run pytest
uv build

Install or refresh all three public Gmail skills:

.\scripts\install-user-skill.ps1

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured