Claude Code Router

Claude Code Router is an open-source Node.js CLI tool designed to enhance Claude Code by allowing users to route coding requests to various AI models. This community-driven extension adds flexibility and customization to AI coding workflows, enabling users to switch between different AI models seamlessly. Claude Code Router supports a variety of providers and models, including OpenRouter, DeepSeek, Ollama (for local models), Google Gemini, and Volcengine. Users can configure default models, background tasks, thinking/reasoning models, and models for long contexts via a JSON file. Switching models on the fly is easy with commands like/model provider,modelwithin the Claude Code interface. Specialized model roles allow users to optimize for different tasks, balancing speed, cost, and capability. Claude Code Router also supports integration with GitHub Actions, streamlining development workflows and automated code reviews or issue bots using any configured model. It is open-source and released under the MIT License, making it free to use and modify. To install Claude Code Router, users need to install Node.js and npm, then install Anthropic’s Claude Code CLI and Claude Code Router. Finally, start it withccr code. By allowing users to route requests to more cost-effective models for specific tasks, it can significantly reduce the overall cost of AI coding assistance compared to using a single, expensive model. Claude Code Router acts as a flexible intermediary, allowing users to route Claude Code requests to different underlying models and customize interactions to their precise needs. Its core capability is the ability to direct Claude Code requests to various models, meaning users are not locked into a single solution and can experiment with or switch to different LLMs based on performance, cost, or specific requirements. A standout feature is its robust plugin system. Developers can create and integrate plugins to rewrite Claude Code prompts and implement custom routing logic. This offers an unparalleled level of customization, enabling highly tailored AI experiences. Whether users need to pre-process inputs or post-process outputs, the plugin architecture makes it possible. One of the most compelling advantages of Claude Code Router is its demonstrated ability to significantly reduce token costs. By seamlessly integrating with more cost-effective models like DeepSeek, users can perform complex AI tasks—such as resolving coding issues—at a fraction of the price of premium models like Claude Max. The project highlights successful real-world scenarios where DeepSeek, via the router, delivered comparable or even superior results in debugging Flutter issues at an astonishingly low cost. Beyond routing and customization, Claude Code Router also supports scheduled tasks, adding another layer of automation and efficiency to users' AI development workflow. The router helps mitigate common pain points, such as 'exceeding context' errors often encountered with certain LLMs. By providing tips on leveraging models with larger context windows (e.g., ByteDance’s DeepSeek API with 128K context), it ensures more stable and consistent long-form conversations. Setting up Claude Code Router is straightforward. After installing Claude Code, users can install the router globally via npm. Then, simply start Claude Code with the router. For plugin integration, users can move their custom plugin files to $HOME/.claude-code-router/plugins and enable them in their config.json file. The repository provides demo plugins for custom routing and prompt rewriting, making it easy to see the system in action. Claude Code Router represents a significant step forward for developers looking to build robust, efficient, and cost-aware AI applications. By offering granular control over AI model interactions and promoting the use of powerful, yet affordable, alternatives, it democratizes access to advanced AI capabilities. This project is a testament to the power of open source in driving innovation and providing practical solutions for real-world challenges in AI development.