Shortcuts MCP Server, developed by Artem Novichkov, connects AI assistants to the Apple Shortcuts ecosystem on macOS to enable AI-driven control of native workflows. The server exposes standardized MCP tools so compatible clients can discover, trigger, and open shortcuts from a chat interface, including passing input data. It is built in Swift for native integration and targets macOS power users, developers, and automation enthusiasts who want AI-triggered workflows inside their existing Shortcuts setups.

What tasks can you actually use it for?

The server turns Shortcuts into discoverable MCP tools that AI clients can call. Its primary outputs are tool entries that let an AI list available shortcuts, initiate a named shortcut with input data, and open a shortcut in the native Shortcuts app for inspection. Compatible clients such as Claude Desktop and Cursor see those functions as standard MCP tools, making it possible to trigger existing macOS workflows directly from a conversational interface.

How reliable are shortcut executions and outputs?

Execution runs on the host using macOS APIs, which keeps invocation local and consistent with native behavior. The Swift-based implementation invokes the same system services the Shortcuts app uses, so run commands execute on the machine rather than being processed remotely. The AI can access shortcut names and trigger them, however the internal logic of a shortcut remains managed by the Shortcuts app itself, so outputs should be validated against the actual workflow results.

What inputs and system requirements determine success?

The tool requires recent system software and a build step for source installs. It needs macOS 14.5 or later to operate and Xcode 16.x to compile from source, with precompiled binaries usable when available. Interoperability depends on clients supporting the Model Context Protocol, and configuration uses simple file edits to point a client at the server binary, so the environment and client compatibility determine whether the integration will work.

Is the setup and workflow fit appropriate for non-technical users?

Integration favors technical users who can modify configuration files and build from source. Typical installation paths involve cloning the Git repository, building with Xcode, and adding the server path to a client's MCP configuration file such as claude_desktop_config.json. The project is open-source, encouraging code inspection and community contributions, and community listings highlight it as an example of macOS-specific MCP integration rather than a plug-and-play consumer tool.

A pragmatic, developer-oriented bridge for AI-driven macOS automation

The server is a practical choice for developers and power users who want local, inspectable automation and are comfortable with a build-and-configure workflow. Expect to treat AI-initiated runs as triggers rather than substitutes for manual testing, and keep Shortcuts logic under local control when validating outputs. For users seeking a development-grade integration between AI clients and macOS workflows, the tool delivers a focused, transparent bridge.