MCP

Written By Manos Mastorakis

Last updated About 18 hours ago

Routine's MCP (Model Context Protocol) support allows external AI tools to connect directly to your Routine workspace. This means tools like ChatGPT Desktop, Claude Desktop, Cursor, and Codex can read your tasks, events, projects, and databases — and act on them — without you switching apps.

Routine supports both an MCP Server (Routine as a data source for external AI tools) and MCP Clients (installing third-party MCP servers to let Routine's AI interact with external services).

✍️ Functionality

With MCP support, you can:

  • Connect external AI tools (ChatGPT Desktop, Claude Desktop, Cursor, Codex) to your Routine workspace

  • Let external AI query your tasks, events, projects, and databases in real time

  • Install third-party MCP servers to give Routine's AI access to external services

  • Use Routine as the central hub for AI actions across multiple tools

  • Keep data local and secure — the MCP server runs on your machine, not in the cloud

  • Link MCP servers in AI prompts using the @ menu and integration pills

  • Manage all connected MCP servers from the AI Integrations tab in Settings

MCP works in both directions — Routine can be a source for external AI, and external services can be sources for Routine's AI.

⁉️ Why use MCP

Most AI tools operate in isolation — they don't know what's on your calendar, what tasks are due, or what projects you're running. MCP changes that by:

  • Giving external AI tools live access to your Routine workspace context

  • Allowing Routine's AI to interact with services it doesn't natively integrate with

  • Keeping everything local and private — the MCP server runs on your computer, not in the cloud

  • Making Routine the connective layer between your AI tools and your productivity data

  • Enabling AI actions that span multiple tools in a single instruction

This is especially powerful for developers and power users who work across many AI tools and want them all grounded in the same workspace context.

🎙️ How to set up MCP

Setting up the Routine MCP Server (Routine → External AI tools):

  1. Open Routine Settings

  2. Navigate to Plugins

  3. Enable the MCP Server plugin

  4. Restart Routine

  5. Follow the setup instructions for your AI tool of choice:

    • ChatGPT Desktop — configure Routine as an MCP source in ChatGPT's settings

    • Claude Desktop — add Routine's MCP server to Claude's configuration file

    • Cursor — connect via Cursor's MCP settings panel

    • Codex — configure via the Codex integration settings

  6. Once connected, your AI tool will list Routine as an available data source

  7. Ask it anything — "What tasks do I have today?" or "What's on my calendar this week?"

💡 The MCP server runs locally on your machine. This means it works with desktop AI tools but not cloud-based ones — and your data never leaves your device.

Setting up MCP Clients (External services → Routine's AI):

  1. Open Routine Settings

  2. Navigate to Integrations → AI

  3. Browse the catalog of available MCP packages

  4. Click to install a vetted MCP server (e.g. a GitHub MCP, a Notion MCP, etc.)

  5. Follow the setup screen for that package

  6. Once installed, Routine's AI assistant, agents, and automations can interact with that service

  7. Reference installed MCP servers in AI prompts using the @ menu or integration pills

Using MCP in practice:

  1. Open any external AI tool (e.g. Claude Desktop)

  2. The tool connects to Routine's MCP server automatically

  3. Ask it to list your tasks, check your calendar, or create a new object

  4. Watch it use Routine's tools to fetch and act on real data

  5. Results appear in both the AI tool and your Routine workspace

🗣️ Use cases

MCP is especially useful for:

  • Developers — use Cursor or Codex to create tasks, log bugs, and check project status without leaving the IDE

  • AI power users — ground ChatGPT or Claude in your real workspace data for more accurate and actionable responses

  • Automated workflows — let external AI tools act on Routine data as part of broader automation pipelines

  • Cross-tool context — ask any connected AI tool "what do I have to do today?" and get a real answer from your Routine workspace

  • External service integration — install MCP servers for GitHub, Notion, or other tools so Routine's AI can interact with them natively

  • Privacy-conscious users — keep all AI interactions local with the on-device MCP server

MCP turns Routine into the central nervous system for all your AI tools — grounded in real data, running locally, and connected to everything.