# Clone and build
git clone https://github.com/adhikjoshi/macpilot.git
cd macpilot
swift build -c release

# Build .app bundle (recommended — needed for Screen Recording permission)
bash scripts/build-app.sh

# Install to /Applications
cp -R MacPilot.app /Applications/
# Symlink for CLI access
ln -sf /Applications/MacPilot.app/Contents/MacOS/MacPilot /usr/local/bin/macpilot

macpilot ax-check --json
# Should show: "trusted": true

# 1. Take screenshot
macpilot screenshot /tmp/screen.png

# 2. OCR the full screen — returns screen coordinates for every text element
macpilot ocr /tmp/screen.png --json
# Returns: { "lines": [{ "text": "Submit", "screenCenterX": 450, "screenCenterY": 300, ... }] }

# 3. Click the element using screen coordinates
macpilot click 450 300

# OCR a specific screen region (x y width height) — faster and more accurate
macpilot ocr 0 30 2240 30 --json    # OCR just the tab bar
macpilot ocr 150 90 150 250 --json  # OCR a sidebar

# Returns same screenCenterX/Y coordinates ready for clicking

macpilot ocr image.png --language ja --json   # Japanese
macpilot ocr image.png --language zh-Hans --json  # Chinese
macpilot ocr image.png --language de --json   # German

# List all shortcuts for the focused app
macpilot ui shortcuts --json
# Returns: { "shortcuts": [{ "shortcut": "cmd+O", "title": "Open File…", "menuPath": "File" }, ...] }

# Use the shortcut directly
macpilot key "cmd+o"     # Open file dialog
macpilot key "cmd+t"     # New tab
macpilot key "cmd+l"     # Focus address bar

# Find text on screen and click it
macpilot screenshot /tmp/s.png
macpilot ocr /tmp/s.png --json | python3 -c "
import sys, json
d = json.load(sys.stdin)
for l in d['lines']:
    if 'Submit' in l['text']:
        print(l['screenCenterX'], l['screenCenterY'])"
# Then: macpilot click <x> <y>

# Find element by label
macpilot ui find "Submit" --json

# Find elements near a position (great for icon buttons without text)
macpilot ui elements-at 340 1200 --radius 50 --json
# Returns all AX elements near that point with roles, actions, and positions

# Click by accessibility label
macpilot ui click "Submit"

macpilot click 100 200             # left click at coordinates
macpilot doubleclick 100 200
macpilot rightclick 100 200
macpilot move 300 300
macpilot drag 100 200 300 400      # drag from (100,200) to (300,400)
macpilot scroll up 5
macpilot scroll down 10
macpilot mouse-position --json     # get current cursor position

macpilot type "Hello World" --json
macpilot key "cmd+c" --json
macpilot key return --json
macpilot key "cmd+shift+3" --json
macpilot keyboard type "text" --json
macpilot keyboard key "ctrl+right" --json   # switch Space

macpilot screenshot /tmp/screen.png --json
macpilot screenshot /tmp/region.png --region 100,100,800,600 --json
macpilot screenshot /tmp/screen.png --with-permissions --json  # use CGWindowListCreateImage directly

# From background processes:
open -n -W -a MacPilot.app --args screenshot --output /tmp/screen.png --json

macpilot app list --json                     # list running apps
macpilot app open Safari --json              # open by name
macpilot app open com.apple.TextEdit --json  # open by bundle ID
macpilot app focus Chrome --json             # bring to front
macpilot app frontmost --json                # get frontmost app
macpilot app quit Safari --json              # graceful quit
macpilot app quit Safari --force --json      # force quit

macpilot window list --json
macpilot window focus Safari --json
macpilot window move Safari 100 100 --json
macpilot window resize Safari 1200 800 --json
macpilot window minimize Safari --json
macpilot window fullscreen Safari --json
macpilot window close Safari --json

macpilot ui list --json                    # list UI elements
macpilot ui find "Submit" --json           # find element by name
macpilot ui click "Submit" --json          # click by accessibility label
macpilot ui tree --depth 3 --json          # element hierarchy
macpilot ui find-text "Search" --json      # search entire AX tree for text
macpilot ui wait-for "Submit" --timeout 10 --json  # poll until element appears
macpilot ui elements-at 340 1200 --radius 50 --json  # find elements near coordinates
macpilot ui shortcuts --json               # list all keyboard shortcuts for focused app
macpilot ui shortcuts --app Chrome --json  # list shortcuts for specific app

macpilot ocr scan /tmp/screen.png --json         # extract text from image with coordinates
macpilot ocr scan 100 100 800 600 --json         # extract from screen region (x y w h)
macpilot ocr scan image.png --language ja --json  # custom language
macpilot ocr click "Submit" --json               # find text on screen and click it
macpilot ocr click "OK" --app Finder --timeout 5 --json  # click text in specific app

macpilot chrome open-url "https://example.com" --json
macpilot chrome new-tab "https://example.com" --json
macpilot chrome list-tabs --json
macpilot chrome close-tab --json
macpilot chrome extensions --json
macpilot chrome dev-mode --json

# Introspect dialog structure (see all fields, buttons, sheets)
macpilot dialog inspect --json

# Navigate to folder in open/save dialog (activates app, uses AX + Go To sheet)
macpilot dialog navigate /tmp --json
macpilot dialog navigate /Users/admin/Desktop --json

# Select a file (without auto-confirming)
macpilot dialog select myfile.txt --json
macpilot dialog select myfile.txt --confirm --json  # select + press Open

# List files visible in dialog
macpilot dialog list-files --json

# Set any text field value in dialog
macpilot dialog set-field "/path/to/file" --json
macpilot dialog set-field "filename" --focused --json

# Click any button in dialog by label
macpilot dialog click-button "Open" --json
macpilot dialog click-button "Cancel" --json

# Trigger file open/save dialogs
macpilot dialog file-open /path/to/file --json
macpilot dialog file-save /path/to/dest --json

# Modal dialog detection & handling
macpilot dialog detect --json              # detect if modal dialog is showing
macpilot dialog dismiss "Don't Save" --json  # dismiss by button name
macpilot dialog auto-dismiss --json        # smart auto-dismiss (Don't Save > OK > Cancel)

# Wait for any dialog to appear across all apps
macpilot dialog wait-for --timeout 30 --json
macpilot dialog wait-for --app Safari --json

# Auto-click the primary/default button of a dialog
macpilot dialog click-primary --json
macpilot dialog click-primary --app Finder --json

macpilot chain "cmd+l" "type:https://google.com" "return" --json
macpilot chain "cmd+l" "type:url" "sleep:500" "return" --delay 200 --json

# Syntax: "key_name", "cmd+key", "type:text", "sleep:ms"

macpilot clipboard get --json              # read text from clipboard
macpilot clipboard set "hello" --json      # set clipboard text
macpilot clipboard image /tmp/clip.png --json  # copy image to clipboard
macpilot clipboard info --json             # content type, size, preview
macpilot clipboard types --json            # list all UTI types on clipboard
macpilot clipboard clear --json            # clear clipboard
macpilot clipboard paste --json            # simulate Cmd+V
macpilot clipboard copy file1.txt file2.pdf --json  # copy files to clipboard
macpilot clipboard save /tmp/saved.png --json       # save clipboard to file

# Clipboard history (background daemon)
macpilot clipboard history start --json              # start tracking (max 50 items)
macpilot clipboard history stop --json
macpilot clipboard history list --json               # show history
macpilot clipboard history list --limit 5 --json     # last 5 items
macpilot clipboard history search "keyword" --json   # search history
macpilot clipboard history clear --json              # delete all history

macpilot shell run "ls -la" --json
macpilot shell run "whoami" --json

macpilot indicator start --json            # start border glow + menu bar icon
macpilot indicator stop --json
macpilot indicator flash --json            # single flash
macpilot indicator status --json           # check if running
macpilot menubar start --json              # ensure menu bar is running (via indicator)

macpilot menubar click Chrome "File > New Window" --json  # click app menu items

macpilot notification send "Title" "Body text" --json   # send system notification
macpilot notification list --json                        # list visible notifications
macpilot notification click --title "Download" --json    # click matching notification
macpilot notification dismiss --json                     # dismiss top notification
macpilot notification dismiss --all --json               # dismiss all notifications

macpilot audio volume --json               # get current volume
macpilot audio volume 50 --json            # set volume (0-100)
macpilot audio mute --json
macpilot audio unmute --json

macpilot display brightness get --json
macpilot display brightness set 0.7 --json  # 0.0-1.0
macpilot display-info --json               # display resolution, refresh rate

macpilot appearance dark --json
macpilot appearance light --json
macpilot appearance toggle --json

macpilot network --json                    # WiFi name, IP, interfaces

macpilot process list --json               # list running processes
macpilot process kill "AppName" --json     # kill by name

macpilot system info --json                # CPU, RAM, disk, OS version

macpilot screen record start --output /tmp/rec.mov --json
macpilot screen record start --output /tmp/rec.mov --region 0,0,1920,1080 --json
macpilot screen record start --output /tmp/rec.mov --window Safari --json
macpilot screen record start --output /tmp/rec.mov --audio --quality high --fps 60 --json
macpilot screen record stop --json
macpilot screen record status --json       # check if recording
macpilot screen record pause --json        # pause recording
macpilot screen record resume --json       # resume recording

macpilot dock show --json
macpilot dock hide --json
macpilot dock autohide --json

macpilot space list --json
macpilot space switch right --json
macpilot space switch left --json
macpilot space switch 1 --json             # by index

macpilot wait element "Submit" --timeout 10 --json
macpilot wait window "Chrome" --timeout 10 --json
macpilot wait seconds 1.5 --json

macpilot login-items list --json

macpilot watch /path/to/dir --json         # watch for file changes

macpilot ax-check --json          # verify Accessibility permission
macpilot --version                # show version (0.7.0)

MP=macpilot

# Example: Navigate Chrome to a URL
$MP app focus "Google Chrome" --json       # PRE: ensure focus
$MP ui shortcuts --json                     # PRE: learn available shortcuts
$MP chain "cmd+l" "type:https://example.com" "return" --json  # ACT
sleep 2
$MP screenshot /tmp/result.png              # POST: verify
$MP ocr /tmp/result.png --json              # POST: read what loaded

# Find and click a specific tab in Chrome
$MP screenshot /tmp/s.png
$MP ocr 0 30 2240 30 --json  # OCR just the tab bar strip
# Parse output to find tab text + coordinates, then click

# Handle icon buttons (no text)
$MP ui elements-at 340 1200 --radius 50 --json
# Find AXGroup/AXButton with actions, get centerX/centerY, then click

# Upload a file in any app (complete workflow)
$MP ui elements-at 340 1223 --radius 30 --json     # Find upload button
$MP click 340 1223 --json                           # Click it
sleep 0.8
$MP screenshot /tmp/menu.png && $MP ocr /tmp/menu.png --json  # Find "Upload a File"
$MP click 395 1065 --json                           # Click menu item
sleep 1.5
$MP dialog detect --json                            # Verify dialog opened
$MP dialog navigate /Users/admin/Desktop --json     # Navigate to folder
$MP click 520 590 --json                            # Click a file
$MP dialog click-button "Open" --json               # Confirm selection

# Debug a dialog (inspect its structure)
$MP dialog inspect --json                           # See all fields, buttons, sheets
$MP dialog set-field "/custom/path" --focused --json  # Set text in focused field
$MP dialog list-files --json                        # List visible files

$MP dialog detect --json                   # check for modal
$MP dialog auto-dismiss --json             # smart dismiss
# Priority: Don't Save > OK > Cancel

swift build                                    # debug build
swift build -c release                         # release build
bash scripts/build-app.sh                      # build .app bundle with ad-hoc signing