How to Automate Your Mac with AI

Your Mac can do more than you ask it to. With AI-powered automation, you can describe tasks in plain English and let the AI figure out the commands, scripts, and file operations. Here is how to set it up and start automating.

What You Need

• A Mac running macOS 14 (Sonoma) or later
• Chapeta installed (free download)
• An OpenRouter API key or Chapeta Pro subscription

Chapeta includes 9 built-in tools that interact with your system. The most powerful automation building blocks are Bash (terminal commands), File Tools (read, write, edit), and Web Search for current context. Together, they cover most practical workflows.

Step 1: Simple File Operations

Start with basic file management. Open Chapeta and try:

“Move all screenshots from my Desktop to ~/Documents/Screenshots”

Chapeta will use its Bash tool to list files matching the screenshot pattern, create the destination folder if needed, and move them. You see each command before it runs and approve it.

More examples:

• “Delete all .DS_Store files in my Documents folder”
• “Create a folder structure for a new project called ‘webapp’ with src, tests, and docs directories”
• “Find all files larger than 100MB on my Desktop”

Step 2: Text Processing and Data

The Bash tool combined with file read/write tools handles data processing:

“Read the CSV file at ~/data/sales.csv, count rows per region, and create a summary”

Chapeta reads the file content, processes it with shell commands like awk or sort, and presents the results. For complex analysis, it can write a Python script and execute it.

More examples:

• “Extract all email addresses from ~/Documents/contacts.txt”
• “Convert all markdown files in this folder to plain text”
• “Find and replace ‘old-domain.com’ with ‘new-domain.com’ in all HTML files”

Step 3: Turn Results into Real Files

The file tools let you turn one-off answers into reusable artifacts:

“Search for the latest OpenRouter pricing docs, summarize the key points, and save the result as ~/Notes/openrouter-pricing.md”

Chapeta can fetch the source material, synthesize it, and write the finished result to disk for you.

More examples:

• “Read every markdown file in this folder and combine the action items into one summary note”
• “Search the web for the latest release notes for Astro and write a quick upgrade checklist to ~/Notes”
• “Find all TODO comments in this repo and save them as a markdown task list”
• “Read this error log and write a short incident summary to my Desktop”

Step 4: Chaining Tools

The real power comes from chaining multiple tools in a single request:

“Take a screenshot of my current screen, describe what you see, and write a summary note to ~/Notes/ui-review.md”

This chains the Screenshot tool (capture), AI analysis (describe), and File Write (save the result). Chapeta handles the orchestration automatically.

“Search my codebase for TODO comments, list them all, and create a markdown file with the results”

This chains Grep (search), file read (context), and file write (output).

Step 5: Save Repeatable Workflows as Skills

If you find yourself running the same automation regularly, save it as a custom skill. Skills are reusable prompts with tool access that you can invoke with a single command.

For example, a “morning briefing” skill could:

1. Check weather via web search
2. Pull the top headlines for your niche
3. Write the summary to a markdown note
4. Present a summary

Once saved as a skill, you run it with one command instead of typing the full prompt each time.

Tips for Effective Automation

• Be specific: “Move PDFs from Desktop” is better than “clean up my Desktop”
• Start small: Test with a single file before running batch operations
• Read the confirmation: Always review the commands Chapeta wants to execute before approving
• Use absolute paths: ”~/Documents/project” is clearer than “the project folder”
• Iterate: If the first result is not right, describe what went wrong and Chapeta will adjust

Limitations

Chapeta cannot run continuously in the background like a cron job. It requires your approval for each tool execution unless you relax the safety mode, so it is interactive automation, not set-and-forget. For unattended scheduled tasks, you still need traditional shell scripts or other schedulers.