Why Connect Your AI to OmniFocus?

If you run your life through OmniFocus, you already take productivity seriously. You have projects, contexts, defer dates, review cycles — the full GTD system. But there is one thing OmniFocus does not do: think with you. It captures and organizes, but it does not help you decide, prioritize, or process.

With LMCP, your AI assistant can read and write your OmniFocus database directly. Ask Claude to look at your inbox and help you process 23 items into projects. Ask it to flag projects with no next action during your weekly review. Tell it to add a task while you are mid-conversation about a feature you want to ship next week.

Your AI becomes the thinking partner OmniFocus was missing — without giving up privacy, without paying for cloud sync of a tool you already paid for, and without learning yet another integration.

What You Will Need

macOS 12 (Monterey) or later — Apple Silicon or Intel
OmniFocus 3 or 4 — the standard Mac app, any license tier (Pro features not required)
Any MCP-compatible AI client: Claude Desktop, Cursor, VS Code (Copilot), Windsurf, Zed, or ChatGPT

You do not need API keys, OAuth tokens, OmniFocus URL automation scripts, or any third-party sync service. LMCP talks directly to OmniFocus on your Mac through its native automation interface, the same one Shortcuts and Automator use.

Step 1: Install LMCP

Open Terminal and run:

curl -fsSL https://local-mcp.com/install?ref=guide-omnifocus-mac | bash

The installer downloads LMCP, creates a background service so it starts automatically, and auto-configures every AI client it detects (Claude Desktop, Cursor, Windsurf, VS Code, ChatGPT). The whole thing takes about 30 seconds.

On first launch, macOS will ask you to grant permission for LMCP to control OmniFocus. Click Allow — this is what lets your AI read your projects and create new tasks through the OmniFocus app.

Step 2: Restart Your AI Client

Quit your AI client completely (Cmd+Q, not just closing the window) and reopen it. It needs to restart to discover the new MCP server.

Claude Desktop — look for the hammer icon in the chat input; the OmniFocus tools should be listed
Cursor — MCP tools appear automatically in the composer
VS Code — check the MCP panel in Copilot settings
ChatGPT — add https://local-mcp.com/mcp as a custom connector

Step 3: Try It Out

Start a new conversation in your AI client and try these prompts:

“List my OmniFocus projects and flag any that have no next action” — the classic stalled-project check from your weekly review
“Show me everything in my OmniFocus inbox and suggest a project for each item” — AI-assisted inbox processing
“Add a task to my Work project: Review Q3 budget proposal, due Friday” — capture without switching apps
“What are my most important tasks today based on flags and due dates?” — daily planning in plain English
“Find every task tagged @errands and group them by location” — context-aware task search

Your AI will always show a preview before creating or completing tasks, so you stay in control of every change to your trusted system.

GTD Workflows That Actually Work With AI

Inbox Processing

The GTD weekly review starts with processing your inbox — the part that most people skip because it is tedious. Instead of doing it alone, share the load:

“Look at my OmniFocus inbox and suggest a project and tag for each item”
“Which inbox items are reference material I should file, not actionable tasks?”
“Process my inbox: for each item, suggest whether it is a next action, a project, a someday/maybe, or should be deleted”

Claude reads every item, applies GTD logic, and gives you a structured recommendation. You make the final calls, but the cognitive load of pattern-matching is shared.

Weekly Review

The weekly review is where GTD lives or dies, and it is where most people fall off the wagon. Claude can make it conversational instead of mechanical:

“Show me all my active projects and flag any that have no next action”
“Which projects have not had any activity in the past two weeks?”
“List my tasks due this week, grouped by project, and tell me which days look overcommitted”
“What am I overcommitted on? Show me days with more than 5 due items”
“Find any projects tagged @waiting-for and tell me which ones I should follow up on”

A weekly review that took 45 minutes becomes a 15-minute conversation with sharper decisions, because you are not just rereading the same project list — your AI is highlighting the things that need attention.

Capture Mid-Conversation

One of the most underrated features: when ideas come up while you are talking with Claude about something else, capture them directly without breaking flow:

“Add a task to my Work project: Review Q3 budget proposal, due Friday, tag as @office”
“Create a task in my inbox: Call the dentist to reschedule, tag as @errands”
“Add these three action items from our conversation to OmniFocus under the Launch Plan project”

No switching apps, no losing your thought, no manual data entry. Claude creates the task in OmniFocus and confirms the details so you know exactly what landed where.

Daily Planning

Start your day with Claude as your planning partner instead of a blank to-do list:

“What are my most important tasks today based on due dates and flags?”
“I have 4 hours of focused time this morning. Which tasks should I prioritize?”
“Show me everything tagged @office — I am heading in today”
“What is the smallest valuable thing I could ship today before lunch?”

The OmniFocus Tools Claude Can Use

LMCP exposes 6 OmniFocus tools through the Model Context Protocol, all read/write and all 100% local. Claude calls these automatically when it needs to interact with your system — you do not call them by name, you just describe what you want.

list_omnifocus_projects — lists your active projects, optionally filtered by folder or tag
list_omnifocus_tasks — lists tasks, with filters for project, tag, due date, flag, or completion state
list_omnifocus_tags — lists your tags so Claude can suggest the right ones during inbox processing
search_omnifocus_tasks — full-text search across task names and notes
create_omnifocus_task — creates a task in any project with name, note, due date, tags, defer date
complete_omnifocus_task — marks a task as done by name or ID

Why Not Just Use Reminders?

Apple Reminders is great for grocery lists and quick captures, but it does not have the GTD primitives you actually use in OmniFocus — folders, defer dates, multi-tag, review intervals, sequential vs parallel projects, perspectives, or the contexts model. If you have ever tried to run a serious GTD system in Reminders, you know it falls apart around 50 active projects.

LMCP supports both, by the way. If you have some things in Reminders (shopping lists, family stuff) and everything else in OmniFocus (work, GTD, projects), Claude can read both and you can ask things like “move this from Reminders to OmniFocus under my Work folder” in plain English.

Privacy: Your Tasks Stay on Your Mac

LMCP runs entirely on localhost. Your OmniFocus database is read locally through OmniFocus's native automation interface — the same way Shortcuts and Automator access it. No task data is sent to external servers beyond what your AI client normally processes during a conversation.

There is no cloud sync layer, no third-party database, no account to create. The data flow is direct: OmniFocus has the data, LMCP reads it, your AI processes it. If you disconnect from the internet, LMCP still works because it talks to OmniFocus, not to any remote service.

For GTD practitioners whose trusted system contains everything — personal, professional, sensitive — this local-only architecture means you can use AI assistance without creating new privacy exposure.

What Else You Get With LMCP

OmniFocus is one piece of a larger system. LMCP ships with 87 tools across 16 Mac applications, so you can ask Claude things like “Check my email for action items from the design team and add them to my Design Review project in OmniFocus” and watch it chain Mail.app reads with OmniFocus writes in a single conversation.

The full surface includes:

Mail — read, search, send, reply from Mail.app
Calendar — list events and create new ones
Contacts — search and read
Microsoft Teams — read chats and channels without Graph API
OneDrive — browse, read, write, and search files
Notes, Reminders, Messages, Safari, Finder — native macOS apps
Office — create and read Word, Excel, PowerPoint, and PDF files

For GTD practitioners, this is the missing piece: an AI that can see your wholesystem — tasks, calendar, email, communication — and help you work it as one.

Get Started

LMCP is free for the first 500 installs. After that, it is a one-time purchase with a 14-day free trial. Install in one command:

curl -fsSL https://local-mcp.com/install?ref=guide-omnifocus-mac | bash

Works with Claude Desktop, Cursor, VS Code (Copilot), ChatGPT desktop, Windsurf, and Zed. Grant the OmniFocus permission when macOS prompts you, restart your AI client, and start asking Claude about your tasks.