@openrouter/cli

OpenRouter CLI

A powerful command-line interface for OpenRouter, enabling seamless integration with multiple language models and AI services. Currently featuring Claude Code integration with more capabilities coming soon.

✨ Features

• Multi-Model Support: Access any model available on OpenRouter through a unified interface
• Claude Code Integration: Use any OpenRouter model with Claude Code for enhanced development workflows
• Smart Model Routing: Automatically route requests to optimal models based on task type (reasoning, long context, web search, etc.)
• Dynamic Model Switching: Change models on-the-fly using simple commands
• Extensible Architecture: Built to support future OpenRouter features and integrations

🚀 Getting Started

1. Installation

Install the OpenRouter CLI:

npm install -g @openrouter/cli

For Claude Code integration, ensure you have Claude Code installed:

npm install -g @anthropic-ai/claude-code

2. Configuration

Create and configure your ~/.openrouter/config.json file. For more details, see the included config.example.json.

The configuration file supports several key sections:

• OPENROUTER_API_KEY: Your OpenRouter API key for accessing models
• LOG (optional): Enable/disable logging. Default is true
• LOG_LEVEL (optional): Set logging verbosity: "fatal", "error", "warn", "info", "debug", "trace". Default is "debug"
• Logging Systems: The OpenRouter CLI uses two separate logging systems:
  • Server logs: HTTP requests and API events in ~/.openrouter/logs/
  • Application logs: Routing decisions and operations in ~/.openrouter/router.log
• models: Configure model routing for different task types
• API_TIMEOUT_MS: API call timeout in milliseconds

Environment Variable Support

The OpenRouter CLI supports environment variable interpolation for secure credential management. Reference environment variables in your config using $VAR_NAME or ${VAR_NAME} syntax:

{
  "OPENROUTER_API_KEY": "$OPENROUTER_API_KEY"
}

Keep sensitive credentials secure by storing them as environment variables rather than in configuration files.

Example Configuration

{
  "OPENROUTER_API_KEY": "$OPENROUTER_API_KEY",
  "LOG": true,
  "models": {
    "default": "deepseek/deepseek-chat",
    "background": "qwen/qwen3-30b-a3b",
    "think": "deepseek/deepseek-reasoner",
    "longContext": "google/gemini-2.5-pro-preview",
    "longContextThreshold": 60000,
    "webSearch": "google/gemini-2.5-flash:online"
  }
}

3. Usage

Claude Code Integration

Use any OpenRouter model with Claude Code:

openrouter code "your prompt here"

Service Management

openrouter start    # Start the proxy service
openrouter stop     # Stop the proxy service
openrouter restart  # Restart the service (required after config changes)
openrouter status   # Check service status

4. Model Routing

Configure intelligent model routing based on task requirements:

• default: Primary model for general tasks
• background: Cost-efficient model for background operations
• think: Advanced reasoning model for complex problem-solving
• longContext: Model optimized for large context windows (>60K tokens)
• longContextThreshold: Token threshold for automatic long-context routing (default: 60000)
• webSearch: Model with web search capabilities (append :online to model name)

Dynamic Model Selection

Switch models on-the-fly within Claude Code:

/model anthropic/claude-sonnet-4

🚧 Coming Soon

The OpenRouter CLI is actively expanding to support:

• Direct model querying/chat
• Ability to intercept all calls on your system to openai or anthropic through your openai account to use any model
• API usage monitoring and analytics
• Account management
• And much more!

Stay tuned for updates as we build out the complete OpenRouter experience.