Automation MCP
An MCP server for macOS: the agent moves the mouse, types, manages windows and captures the screen when an app has no API
- 414 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent controls the mouse and keyboard of the whole computer and can press any key combination, including system shortcuts
- Screenshots show the agent everything open on screen, including passwords and private messages
Install
- Works in
Manual install
furi add ashwwwin/automation-mcp
furi start ashwwwin/automation-mcpRequires the furi manager (github.com/ashwwwin/furi) or its desktop app.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent controls the mouse and keyboard of the whole computer and can press any key combination, including system shortcuts
- Screenshots show the agent everything open on screen, including passwords and private messages
This is third-party code. Review the repository files before installing.
What it does
Automation MCP is an MCP server for macOS that lets an AI model operate the computer. The agent clicks, moves the cursor, scrolls and drags, types text and presses key combinations, including system shortcuts. For the mouse there is path movement through several points as well as separate button press and release. The server returns the screen as screenshots: the full screen, a region or a single window. Besides screenshots it offers screen dimensions, the color of any pixel, region highlighting and waiting for an image to appear by template. Windows are handled through the list of open windows and the active window: focus, move, resize, minimize. The server runs through the furi manager or directly on Bun. HTTP mode is recommended for web apps, and stdio mode suits command-line clients and Claude Desktop. On the first run macOS asks for two permissions: Accessibility and Screen Recording.
Who it is for. For developers and testers who want an agent to work inside a Mac app that has no API: click, type and check the screen.
Good fit when
- The agent has to walk through the interface of a Mac app that has no API: click by coordinates and type text
- You need a screenshot of a window or a screen region for the model to analyze: the `screenshot` tool
- A scenario waits for a button or dialog to appear: `waitForImage` matches the screen against a template
- You need to arrange windows: focus, move, resize and minimize
- You need keyboard shortcuts: `systemCommand` runs copy, paste, undo and save
Not a fit when
- You work on something other than macOS: the server is documented for Mac only
- You cannot grant the process Accessibility and Screen Recording permissions
- A password manager, a bank or private messages are open on screen: the agent sees the whole screen and can click anywhere
- You need work with interface elements by meaning: the server acts by screen coordinates
Example request
Take a screenshot of the Safari window, find the Send button and click it, then wait for the checkmark image to appearMCP
- Transport
- stdio, http
- Authentication
- not required
Limitations
- Works on macOS only and requires the Bun runtime.
- Without the Accessibility and Screen Recording permissions neither mouse and keyboard control nor screenshots work.
- The server locates interface elements by coordinates, so a scenario depends on screen resolution and window layout.
- Running without furi means cloning the repository and installing dependencies by hand, and the client configuration points to `index.ts` on your machine.
How to disable. Remove the automation entry from the MCP client configuration and stop the server process. You can revoke the permissions in System Settings under Privacy & Security, in the Accessibility and Screen Recording sections.
FAQ
How do I connect Automation MCP to Claude Desktop?
Add an automation server to the MCP configuration with command bun and arguments run, the path to `index.ts` in the cloned repository and `--stdio`. If Claude Desktop is already set up through furi, nothing needs to be added.
Which permissions does macOS need?
Two: Accessibility for the mouse, keyboard and window control and Screen Recording for screenshots, screen analysis and color detection. macOS asks for them on the first run, and you can enable them later in System Settings.
How does HTTP mode differ from stdio?
`bun run index.ts` starts the HTTP transport, which the author recommends for web apps. With the `--stdio` flag the server talks to the client through standard input and output, the route for command-line tools and Claude Desktop.
What can the server do with the screen?
The `screenshot` tool captures the full screen, a region or a window, `screenInfo` reports dimensions, `colorAt` returns a pixel color, `screenHighlight` marks a region and `waitForImage` waits for an image to appear by template.
How do I run the server through furi?
Run `furi add ashwwwin/automation-mcp`, then `furi start ashwwwin/automation-mcp`. The furi desktop app offers the same actions without the command line.
Discussion
No questions or reviews yet.
Related
MCP serversA Swift MCP server that lets an agent drive any macOS app: click, type and read the interface tree
An MCP server for driving Windows: the agent opens apps, clicks and types, reads window and UI state, and works with files and PowerShell
MCP server for AppleScript: the agent runs scripts and controls Mac apps, from Notes and Calendar to files
An MCP server for Windows: the agent moves the mouse, presses keys, manages windows and takes screenshots
Open Computer Use
open-computer-use
An open MCP server for desktop control: the agent reads app state and clicks, types and scrolls on macOS, Linux and Windows
A macOS CLI and MCP server: screenshots, accessibility-based UI inspection and app control via clicks and typing
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Automation MCP (this entry) | no | not required | 414 | |
| Windows-MCP | no | not required | 10,740 | the same job for Windows: apps, clicks, typing, window state, files and PowerShell |
| Remote macOS Use: control of a remote Mac | no | API key | 490 | shows Claude Desktop a Mac screen over VNC and controls it remotely |
README in short
The page opens with the feature list and a quick start through furi, then shows the start without it, the tools grouped by area and Claude Desktop connection examples. It ends with the permissions section and troubleshooting tips.