A tool for automatically accepting Cursor's AI suggestions.
The analyze_hover_results.py script is the current working version for analyzing button positions and matches. This script:
- Takes a full screen screenshot
- Analyzes potential button matches using template matching
- Groups matches by x-axis position
- Prioritizes matches by confidence and y-axis position
- Generates a visualization showing calibration points and matches
- Uses color coding to distinguish between different buttons (green, yellow, red)
Other calibration scripts have been moved to the _archive directory.
git clone https://github.com/yourusername/cursor-auto-accept
cd cursor-auto-accept
./setup.sh
This will: - Create necessary directories - Set up permissions - Create a Python virtual environment - Install required packages
Before first use, you need to calibrate the bot for each monitor where you use Cursor:
./stop_clickbot.sh
For all monitors:
bash
source venv/bin/activate
python cursor_auto_accept.py --capture
For a specific monitor (0-based index):
bash
python cursor_auto_accept.py --capture --monitor 0 # First monitor
python cursor_auto_accept.py --capture --monitor 1 # Second monitor
The bot will save separate accept button images for each monitor in:
assets/monitor_0/accept_button.png
assets/monitor_1/accept_button.png
...
./start_clickbot.sh
The bot will: - Check if another instance is running - Load calibration images for all monitors - Start monitoring all screens - Begin accepting prompts automatically
./stop_clickbot.sh
The bot logs all activity to temp/logs/clickbot.log. You can monitor it with:
tail -f temp/logs/clickbot.log
The bot has several built-in settings: - Rate limit: 8 clicks per minute - Confidence threshold: 0.8 (80% match required) - Search interval: 0.2 seconds - Log update interval: 5 seconds
python cursor_auto_accept.py --capture --monitor XEnsure Cursor's accept button is visible on that monitor
If clicks are inaccurate:
Try calibrating in different lighting conditions
If the bot won't start:
temp/clickbot.pid.
├── assets/ # Calibration images
│ ├── monitor_0/ # First monitor
│ │ └── accept_button.png
│ └── monitor_1/ # Second monitor
│ └── accept_button.png
├── temp/ # Runtime files
│ ├── clickbot.pid # Process ID
│ └── logs/ # Log files
├── cursor_auto_accept.py # Main bot script
├── setup.sh # Setup script
├── start_clickbot.sh # Start script
└── stop_clickbot.sh # Stop script
MIT License
$ claude mcp add cursor-jailbreak \
-- python -m otcore.mcp_server <graph>