OpenPal

WebSocket broker for AI CLI tools with session management and real-time streaming.

Features

• Multi-provider support (Claude, Codex, Copilot, Gemini, OpenCode)
• WebSocket API for real-time communication
• Session persistence and resume
• Concurrent client handling
• JSON streaming output

Installation

# Clone the repository
git clone <repository-url>
cd OpenPal

# Build
make build

# Or install to /usr/local/bin
make install

Usage

# Basic usage
./openpal --quest-id my-task --provider claude --work-dir /path/to/project

# With specific port
./openpal --quest-id my-task --provider claude --port :8080

# Show version
./openpal --version

Environment Variables

• PAL_PROVIDER - Default AI provider (claude, codex, copilot, gemini, opencode)
• PAL_WORK_DIR - Working directory for AI CLI
• PAL_<PROVIDER>_PATH - Custom path to CLI executable
• PAL_CAPABILITIES - Comma-separated list of capabilities
• ANTHROPIC_API_KEY - API key for Claude
• OPENAI_API_KEY - API key for Codex
• GITHUB_TOKEN - Token for Copilot

WebSocket API

Connect to ws://localhost:<port>/ws?device=<device-id> to stream AI responses.

Development

# Run tests
make test

# Run tests with coverage
make coverage

# Format code
make fmt

# Run linter
make lint

# Build for all platforms
make build-all

Configuration

Create a .env file in your working directory:

ANTHROPIC_API_KEY=your-key-here
PAL_PROVIDER=claude

License

See LICENSE file.