A Discord-style, voice-first interface for interacting with multiple Cursor AI agents in parallel. Replace tab-based chat UIs with a single collaborative call.

Note: This project is currently in active development. Features may be incomplete or subject to change.

Codecall is a VS Code/Cursor extension that reimagines how developers interact with AI coding assistants. Instead of managing multiple chat tabs, Codecall presents a unified "video call" interface where:

• Each AI agent appears as a tile in a grid layout
• Agents work on tasks in parallel and report back via voice
• You speak to agents using push-to-talk
• Agents queue up to speak, one at a time, like a real meeting

• Spawn multiple agents - Create AI agents that work on different tasks simultaneously
• Visual status indicators - See at a glance which agents are idle, listening, working, or reporting
• Live output streaming - Watch agent output stream in real-time as captions on each tile
• File tracking - Track which files each agent reads and modifies

• Push-to-talk input - Speak naturally to give agents instructions
• Text-to-speech output - Agents summarize their work and speak it back to you
• Speaking queue - Agents take turns speaking; you control who goes next
• Multiple voice presets - Choose from professional, friendly, technical, calm, or energetic voices
• Auto-open files - When an agent speaks, the files it modified automatically open in the editor so you can follow along

• Cursor CLI integration - Spawns agents via the Cursor headless CLI
• ElevenLabs voice services - High-quality text-to-speech and speech-to-text
• Real-time streaming - JSON streaming for live progress updates

• VS Code or Cursor editor
• Node.js v18 or higher
• Cursor CLI (agent command available)
• API Keys:
  • Cursor API Key - For headless agent operations
  • ElevenLabs API Key - For voice features (TTS/STT)

git clone https://github.com/yourusername/codecall.git
cd codecall

npm install

Copy the example environment file and fill in your API keys:

cp .env.example .env

Edit .env with your values:

# Required
CURSOR_API_KEY=your_cursor_api_key_here
ELEVENLABS_API_KEY=your_elevenlabs_api_key_here

# Optional
ELEVENLABS_AGENT_ID=your_agent_id_here
PORT=3000

npm run compile

In a separate terminal:

npm run server

Or with auto-reload during development:

npm run server:watch

• Open the project in VS Code/Cursor
• Press F5 to launch the Extension Development Host
• The Codecall panel will appear in the sidebar

This section provides detailed instructions for running the extension in debug mode during development.

1. Open the project in VS Code or Cursor:

   code "/home/areg_/Code Ubuntu/codecall"
   # or
   cursor "/home/areg_/Code Ubuntu/codecall"

2. Install dependencies (if not already done):

   npm install

3. Start the backend server in a terminal:

   npm run server

4. Launch the Extension Development Host:

   • Press F5 (or go to Run > Start Debugging)
   • This will:
     • Automatically compile the extension via the npm run watch task
     • Open a new VS Code/Cursor window (Extension Development Host)
     • Load the extension in development mode

5. Access Codecall:

   • In the new window, click the Codecall icon in the Activity Bar (left sidebar)
   • The Codecall panel will open

For active development with hot-reloading:

Terminal 1 - Watch for extension changes:

npm run watch

Terminal 2 - Backend server with auto-reload:

npm run server:watch

Terminal 3 (Optional) - Full dev environment:

npm run dev

Then press F5 to launch the Extension Development Host. Changes to TypeScript files will be automatically recompiled.

• Breakpoints: Set breakpoints in .ts files under src/. The debugger will pause at these points.
• Debug Console: View console.log output from the extension in the Debug Console (View > Debug Console).
• Webview DevTools: Right-click the Codecall sidebar panel and select "Developer: Open Webview Developer Tools" to debug the React UI.
• Reload Extension: Press Ctrl+Shift+F5 (or Cmd+Shift+F5 on Mac) to reload the Extension Development Host after making changes.

The debug configuration is defined in .vscode/launch.json:

{
  "name": "Run Extension",
  "type": "extensionHost",
  "request": "launch",
  "args": ["--extensionDevelopmentPath=${workspaceFolder}"],
  "outFiles": ["${workspaceFolder}/out/**/*.js"],
  "preLaunchTask": "${defaultBuildTask}"
}

You can also configure API keys through VS Code/Cursor settings:

1. Open the Codecall sidebar panel
2. Click "Spawn Agent" or run Codecall: Spawn Agent from the command palette
3. Enter a task prompt (e.g., "Refactor the authentication module")
4. The agent appears as a new tile and begins working

• Single-click an agent to select it for voice input
• Double-click a working agent to interrupt it (puts it in listening mode)
• Dismiss an agent when you're done with it

• Hold the push-to-talk button to speak to the selected agent
• When an agent finishes a task, it generates a summary and queues to speak
• Click "Allow to Speak" on a queued agent to let it go next
• Agents speak one at a time to avoid overlapping audio

┌─────────────────────────────────────────────────────────────┐
│                     VS Code Extension                        │
├─────────────────────────────────────────────────────────────┤
│  extension.ts          │  Webview Provider, message routing │
│  agentManager.ts       │  Cursor CLI process management     │
│  voiceManager.ts       │  ElevenLabs TTS/STT integration    │
├─────────────────────────────────────────────────────────────┤
│                     Webview UI (React)                       │
│  App.tsx               │  Main chat interface               │
│  components/           │  UI components                     │
│  hooks/                │  Voice interaction hooks           │
└─────────────────────────────────────────────────────────────┘
           │                            │
           ▼                            ▼
┌──────────────────┐          ┌──────────────────┐
│   Cursor CLI     │          │  Hono Server     │
│   (agent -p)     │          │  (server.ts)     │
│                  │          │                  │
│  • Stream JSON   │          │  • REST API      │
│  • File ops      │          │  • TTS/STT       │
│  • Tool calls    │          │  • Agent state   │
└──────────────────┘          └──────────────────┘

# Build everything
npm run compile

# Watch mode (extension + webviews)
npm run watch

# Run the backend server
npm run server

# Run server with auto-reload
npm run server:watch

# Lint the codebase
npm run lint

# Run tests
npm test

codecall/
├── src/
│   ├── extension.ts        # Extension entry point
│   ├── agentManager.ts     # Agent lifecycle management
│   ├── voiceManager.ts     # Voice services
│   └── webview-ui/
│       └── sidebar/        # React UI components
├── server.ts               # Backend API server
├── package.json            # Extension manifest
└── .env.example            # Environment template

The server exposes the following REST endpoints:

• Basic agent spawning and management
• Cursor CLI integration with streaming output
• ElevenLabs TTS integration
• Speaking queue management
• File tracking (modified/read files per agent)
• Auto-open files when agent reports
• Push-to-talk speech input (STT)
• Agent tile grid UI with waveforms
• Screen sharing for agents (code walkthrough)
• Persistent conversation history
• Multi-workspace support

Contributions are welcome! Please feel free to submit issues and pull requests.

MIT

• Cursor for the AI-powered editor and CLI
• ElevenLabs for voice synthesis
• Vercel AI SDK for AI integrations