macos-automator-mcp
Lets MCP clients discover and run AppleScript or JavaScript for Automation (JXA) to control macOS applications, inspect the system, or reuse scripts from a bundled knowledge base.
README
macOS Automator MCP 🤖 — Give your agent a Mac to operate

macOS Automator MCP is a Model Context Protocol server that lets MCP clients discover and run AppleScript or JavaScript for Automation (JXA). It is for agents that need to control macOS applications, inspect the system, or reuse scripts from a bundled knowledge base.
Install
You need macOS and Node.js 24 or newer. Add the server to your MCP client's configuration; npx downloads the current npm release when the client starts it.
{
"mcpServers": {
"macos_automator": {
"command": "npx",
"args": ["-y", "--package", "@steipete/macos-automator-mcp", "macos-automator-mcp"]
}
}
}
If your client has a separate package field, use @steipete/macos-automator-mcp without @latest.
Quick start
Restart your MCP client after adding the configuration. First, ask it to call get_scripting_tips with a small search:
{
"search_term": "Safari front tab URL",
"limit": 3
}
Then verify script execution with a read-only inline script through execute_script:
{
"script_content": "return \"Hello from macOS Automator\""
}
The result is Hello from macOS Automator. Calls that control applications or the user interface may prompt for macOS permissions.
Tools
| Tool | Purpose |
|---|---|
get_scripting_tips |
List knowledge-base categories or search for AppleScript and JXA tips. |
execute_script |
Run one inline script, script file, or knowledge-base script ID. |
Use get_scripting_tips before writing a script from scratch. A returned runnable ID can be passed to execute_script as kb_script_id; scripts with placeholders accept named input_data or positional arguments.
execute_script runs with the privileges of the process hosting the MCP server. Only run scripts you trust, and inspect generated scripts before allowing destructive actions. See the tool reference for every input and response option.
Permissions
The application that launches the MCP server—such as Terminal, an editor, or a desktop MCP client—owns its macOS privacy permissions:
- Grant Automation access when scripts control Finder, Safari, Mail, or another application.
- Grant Accessibility access when scripts use System Events for clicks, keystrokes, menus, or other UI scripting.
macOS may show a first-use prompt for each target application. The server cannot grant these permissions itself. See configuration and permissions for setup and common error codes.
Knowledge base
The package includes hundreds of AppleScript and JXA tips covering system tasks, files, browsers, terminals, productivity apps, developer tools, and UI automation. Search by keyword or category, then execute a result by its runnable ID.
A local knowledge base can add or override bundled tips without changing the package. It defaults to ~/.macos-automator/knowledge_base; see configuration and permissions for its layout and override rules.
Configuration
| Variable | Values | Default |
|---|---|---|
LOG_LEVEL |
DEBUG, INFO, WARN, ERROR |
INFO |
KB_PARSING |
lazy, eager |
lazy |
LOCAL_KB_PATH |
Absolute path to a custom knowledge base | ~/.macos-automator/knowledge_base |
lazy loads the knowledge base on first use; eager loads it at server startup. More detail is in configuration and permissions.
Troubleshooting
- Permission errors such as
-1743or-10004usually mean the host application needs Automation or Accessibility access. - Script syntax errors are easiest to isolate with
include_executed_script_in_outputandinclude_substitution_logs, then reproduce in Script Editor. - Use an absolute POSIX path with
script_path, and raisetimeout_secondsfor scripts that legitimately need more than 60 seconds. - JXA normally works best with
output_format_mode: "direct"; the defaultautomode selects it for JXA.
See Debugging AppleScript and JXA for a longer diagnostic guide.
Development
pnpm install
pnpm run build
pnpm test
pnpm run lint
pnpm run validate
The repository uses pnpm 11 and Node.js 24. The development guide covers local server setup and knowledge-base contributions.
Community
Report bugs and propose scripts in GitHub Issues.
<a href="https://glama.ai/mcp/servers/@steipete/macos-automator-mcp"> <img width="380" height="200" src="https://glama.ai/mcp/servers/@steipete/macos-automator-mcp/badge" alt="macOS Automator MCP server on Glama" /> </a>
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.