OmniFocus

OmniFocus MCP Server provides a bridge between Claude (or other MCP-compatible clients) and your OmniFocus database. It exposes a set of MCP tools that let you query, add, and manage tasks and projects within OmniFocus through natural language conversations. You can run the server locally and connect it to Claude Desktop so that your AI assistant can list tasks, create new actions from transcripts, bulk edit projects, and generate focused views using the built-in query capabilities. The server also supports resources to preload data contexts (like inbox, today, flagged, and perspective-based views) for faster conversational interactions.

Available tools include query_omnifocus for targeted searches (filter by project, tags, status, due/deferred/planned dates, and more), dump_database for exporting the full OmniFocus state, add_omnifocus_task and add_project for creating items, and remove_item for deleting tasks or projects. You can also work with perspectives and template resources to tailor your results for specific folders, projects, or named perspectives. Use cases range from converting meetings or transcripts into actionable OmniFocus items to generating consolidated task lists and next actions across multiple projects.

Prerequisites:

• macOS with OmniFocus installed
• Node.js and npm installed (to run via npx)

Steps:

1. Install Node.js (if not already installed).

   • macOS (via Homebrew): brew install node
   • Or download from https://nodejs.org/

2. Install and run the OmniFocus MCP server using npx:

   • You only need an active internet connection for the first run:

   npm install -g npm@latest

3. Configure Claude (or your MCP client) to connect to the server:

   • In Claude Desktop, add a new MCP server with the following configuration:

   { "mcpServers": { "omnifocus": { "command": "npx", "args": ["-y", "omnifocus-mcp"] } } }

4. Restart Claude Desktop to load the new MCP server configuration.

Notes:

• The server is started via npx, so an active npm installation is sufficient; you don’t need to run a separate daemon process.
• Ensure OmniFocus is installed locally on macOS where you run the MCP server.

Tips and considerations:

• Known Issues: Dump_database can be slow or fail with very large OmniFocus databases. For large datasets, prefer query_omnifocus for targeted data or incremental exports.
• Resources: Use fixed resources (omnifocus://inbox, today, flagged, stats) and template resources (omnifocus://project/{name}, omnifocus://perspective/{name}) to preload context for faster conversations.
• Tools overview: query_omnifocus supports complex filters and selective fields to minimize payload size. add_omnifocus_task and add_project let you create items with optional due dates, tags, and notes. remove_item can delete tasks or projects by id or name.
• Environment and configuration: The default install uses npx to fetch the omnifocus-mcp package. If you deploy in a container or CI, you may switch to a package build and run via node directly if supported by the package author.
• Security: Treat OMNI focus data as sensitive; restrict MCP access to trusted clients and manage permissions accordingly.