A new MCP server brings Mail, Calendar, Contacts, Teams, and OneDrive to AI assistants — all running locally with zero cloud processing. The privacy-conscious enterprise use case gets a compelling answer.
Most productivity MCP servers require OAuth tokens, API keys, or cloud services. For enterprises and privacy-conscious users, this creates a fundamental tension: you want AI assistance with your email and calendar, but you don't want your data leaving your machine.
Local MCP takes a different approach: everything runs on your Mac, reading from native apps and local data stores. No tokens, no APIs, no cloud — just direct access to your productivity data.
Perhaps most impressive: Teams chat/channel access works by reading Microsoft Teams' local IndexedDB storage. No Graph API tokens required. This is the kind of pragmatic solution that makes self-hosted AI assistants viable for enterprise users.
Read, search, send, reply via Mail.app. Works with Gmail, Outlook, iCloud, Exchange.
List, create, delete events. Multi-account support for all calendar providers.
Search and list contacts from the native Contacts app.
Read chats and channels via local IndexedDB. No Microsoft API needed.
Full file management: read, write, search, move, delete.
Word, Excel, PowerPoint, PDF reading and extraction.
The zero-config approach is notable:
npm install -g local-mcpThat's it. The package auto-detects which AI tools you have installed and configures them automatically. For manual configuration:
{
"command": "npx",
"args": ["-y", "local-mcp@latest"]
}Local MCP includes proper MCP discovery metadata:
https://www.local-mcp.com/.well-known/mcp.jsonBefore sending an email, the tool shows a preview of what will be sent. This "confirm before send" pattern is essential for AI assistants with write access — a single mistaken email could cause real damage.
This safety-first approach aligns with emerging best practices in agentic tooling: tools that can cause harm should require explicit confirmation, especially when the AI might be operating autonomously.
The privacy architecture has several advantages:
The local-only approach does have tradeoffs:
Local MCP joins a growing category of "local-first" MCP servers that prioritize privacy:
The pattern: give AI assistants access to your local environment without requiring cloud services. As AI assistants become more capable, the question of "where does my data go?" becomes increasingly important.
The PR adds Local MCP to the official MCP servers README. With 1,000+ downloads already, adoption is strong. Watch for:
Links:
Published March 28, 2026 · View Repo Pulse · Home