macOS Mail MCP
Discovers and reads mail accounts configured in Apple Mail via macOS automation, allowing MCP clients to access email without storing passwords.
README
macOS Mail MCP
An MCP server that discovers the mail accounts already configured in Apple Mail and reads them through Apple Mail automation. There is no accounts configuration file, and this project never writes a password or token to disk.
Status and macOS boundary
macOS has no supported public API for enumerating the detailed incoming/outgoing server settings of every Internet Account. This server reads the current Sonoma accountd store at ~/Library/Accounts/Accounts4.sqlite (and supports legacy Apple Mail Accounts.plist metadata when present), then obtains credentials only on demand with the macOS security tool. This is deliberately best-effort across Mail versions: test it against your particular Sonoma-or-newer Mail configuration before using it for production workflows.
The first mail operation may show an Automation prompt. Allow the ChatGPT desktop app to control Mail. Do not export/import passwords: Apple Mail performs its own account authentication, including provider OAuth.
Because Apple Mail stores its account database in a TCC-protected Library location, macOS also requires Full Disk Access for the application launching this server. In System Settings → Privacy & Security → Full Disk Access, add Terminal/iTerm if you launch it there; add your MCP desktop client (for example Codex) when that client launches it. Restart the launching application afterwards.
Install
Requires macOS Sonoma+, Python 3.12+, and Apple Mail configured with IMAP-capable accounts.
git clone git@github.com:wherecodegetscooked/applemail-mcp.git
cd applemail-mcp
python3.12 -m venv .venv
.venv/bin/pip install -e '.[dev]'
.venv/bin/pytest
Run it over standard input/output (the normal MCP transport):
.venv/bin/macos-mail-mcp
Register the server with the ChatGPT desktop app / Codex:
codex mcp add macos-mail -- "$PWD/.venv/bin/macos-mail-mcp"
Restart the ChatGPT desktop app, then ask a new Codex task about your mail. Keep stdout reserved for MCP traffic; diagnostics go to stderr/the launchd log.
Architecture
discovery.pyscans macOS account metadata every cycle. Account additions, removals, and server changes are picked up without a server restart.apple_mail.pyuses macOS JXA automation to read the already-synced Apple Mail app. Apple Mail remains responsible for IMAP, SMTP, provider OAuth, retries, and credentials.service.pyqueries Apple Mail directly for current mail and search, then caches returned messages in SQLite/FTS5 for subsequent local reads.store.pycontains message metadata, indexed bodies, and attachment metadata. It has no credentials. Attachment bytes are fetched only afterread_attachment.
Security and operations
The SQLite cache is private to the user at ~/Library/Application Support/macOS-Mail-MCP/mail.sqlite3; protect the macOS login account and use FileVault. It can contain email content, so do not place it in a shared directory. Set MAIL_MCP_DB to another private path if required and MAIL_MCP_LOG_LEVEL=DEBUG for troubleshooting. Logs never include passwords.
Apple Mail performs provider authentication itself, so OAuth-only accounts such as Gmail can be queried without making their tokens available to this server. The server needs Full Disk Access for account discovery and Automation permission to read Mail.
launchd
launchd/com.example.macos-mail-mcp.plist is a template. Replace both REPLACE_ME occurrences and the executable path with the virtualenv’s absolute executable path, then install it:
mkdir -p ~/Library/LaunchAgents
cp launchd/com.example.macos-mail-mcp.plist ~/Library/LaunchAgents/
launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.example.macos-mail-mcp.plist
For an stdio MCP server, launchd is normally used by a client-side bridge; most MCP desktop clients instead start the command themselves. Do not start two instances against the same SQLite file unless you have intentionally chosen that arrangement.
Testing
Unit tests cover plist normalization, Keychain command handling, and SQLite FTS/incremental sync state. Integration testing should use a dedicated disposable IMAP account and an isolated MAIL_MCP_DB; it must never use a personal mailbox in CI. Run pytest locally after configuring that account.
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.