The conversational assistant for vibe coders
Features | Supported AI IDEs | Installation | Current Limitations
SuperCode enhances your coding experience by letting you interact with your IDE using voice commands. Vibe coding how it should be.
-> 🌉 Based in SF? Let's collaborate and shape the future of conversational coding
- Only supports MacOS + English for now
- Requires Accessibility permissions to capture your screen
- After installation, SuperCode appears in your menu bar. Launch it anytime with Command + Option + L (
⌘⌥L), or click the menu bar icon - When running SuperCode for the first time:
- Open your IDE (e.g. Windsurf) on your primary monitor and make sure it's in focus
- Wait for initialization to complete (the "initializing" text in the overlay will disappear)
- Keep your IDE on the primary monitor - multiple monitor support coming soon!
- Voice-Activated Commands: Trigger actions with the "activate" keyword
- IDE Integration: Seamlessly works with supported IDEs (Cursor, Lovable, etc.)
- Real-time Status Display: Always know what SuperCode is doing with the overlay UI
Talk to your IDE with natural language commands:
- "Activate type [text]" - Have the AI type and edit code for you
- "Activate change [ide_name]" - Switch to a different supported IDE
- "Activate learn [element] [name]" - Teach SuperCode about a new UI element
- "Activate click [element]" - Click a learned UI element
- "Activate find [text]" - Search for text in the current file Real-time voice notifications - receive instant alerts when coding tasks are completed, eliminating the need to manually check if the coding agent is done.
- Cursor - The AI-first code editor
- Windsurf - Open source AI coding environment
- Lovable - AI-powered web IDE for full-stack development
- Coming Soon: Claude Code, Base44, Bolt, v0
For the easiest setup experience, use our installation script:
# Clone the repository
git clone https://github.com/saharmor/supercode.git
cd supercode
# Run the installation script (automatically sets up everything)
./install_and_run.shThe script will:
- Check for Python 3.8+ and install requirements
- Create a virtual environment
- Prompt for required API keys if not found
- Configure the .env file (using .example.env as a template)
- Launch SuperCode
- Anthropic API key for Claude Computer Use see guide for getting your key
- OpenAI API key for Whisper transcription, optional, otherwise uses Google free ASR see guide for getting your key
- Google Gemini API key for screenshots and image analysis see guide for getting your key
SuperCode uses a .env file for configuration. An example template (.example.env) is provided with the repository. The installation script will automatically create this file for you, but you can also manually copy and edit it:
cp .example.env .env
# Then edit .env with your preferred text editor- macOS (currently macOS-only)
- Requires Accessibility permissions
- Approximately $0.2/hour with default settings, originating from Claude Computer Use, which finds the location of the interacted elements on yoru screen, and OpenAI
gpt-4o-transcribe - Can be ~80% cheaper using Google's free ASR, but at the cost of lower transcription quality
Activate [command] [parameters]
Example commands:
Activate type Write a function that sorts an array- Start typing in the IDEActivate change lovable- Switch to the Lovable interfaceActivate learn Accept the blue button to the right that runs the suggested terminal code- Teach a new button
SuperCode is actively being developed, and there are some current limitations to be aware of:
- English Only: Currently only supports English language for voice commands and responses
- Display Requirements: The IDE should remain open on your primary display for proper screen monitoring and interaction
- macOS Only: Currently only compatible with macOS, with Windows and Linux support planned for future releases
SuperCode interacts with several external APIs to function. Here's what data is sent:
Screenshots
- When SuperCode starts and when you switch IDEs, screenshots are sent to Claude Computer Use to find screen coordinates for UI elements (input boxes, buttons, etc.)
- After executing commands, screenshots may be sent to Gemini Flash to monitor when the command is completed
- These screenshots are only of your active IDE window and used only for application functionality
Audio
- Your voice commands are sent to either OpenAI API (if enabled) for high-quality transcription or Google's free speech recognition service
- Audio is only transmitted when actively recording a command
All data transmission is done securely via API providers' encrypted connections. No data is stored by SuperCode beyond the current session.
- Electron app for a more native experience
- Claude Code integration with reminder to run /compact
- Consolidating API providers (currently using Claude, Gemini, and GPT)
- Gemini Flash / OpenAI Realtime transcription
- Model selector when using "Activate Type" to optimize credits usage
- Error detection and fixing suggestions
- Rabbit hole detector to keep you on track
- Commit message generation for Windsurf
- Realtime audio APIs for faster and smoother experience
- Function calling for more robust voice command processing
- Support for additional languages: Spanish, German, and French
Are you based in San Francisco? Let's code! Reach out to us on X @theaievangelist.
Apache 2.0