Claudish is a command-line tool that connects various AI models through a local proxy, allowing users to leverage existing AI subscriptions without incurring additional costs. It supports a wide range of models, including those from Anthropic, Google, OpenAI, and local hosting options. Users can run requests through a compliant server, ensuring privacy and flexibility. The tool is especially useful for developers looking to streamline their workflow by consolidating multiple AI services into one interface. The article details how to utilize Claudish with specific commands for different models. For instance, users can run the command `claudish --model g@gemini-3-pro-preview` for Gemini Advanced or `claudish --model oai@gpt-5.3` for ChatGPT Plus. It highlights a 100% offline option where code doesn’t leave the machine, emphasizing security. Claudish also supports multiple providers and allows users to switch models mid-session. The routing syntax lets users specify models directly, enhancing control over the AI interaction. Installation options are straightforward, whether through curl, Homebrew for macOS, or npm. For those who want to use it without installation, npx allows access to the latest version directly. The setup prompts users for API keys, making it user-friendly even for those unfamiliar with command-line interfaces. Claudish also includes features like real-time output streaming, parallel runs, and a headless mode for automated tasks. Key features include auto-detection of models, a monitoring mode for debugging, and JSON output for easy integration with other tools. The article emphasizes the importance of file-based instructions to prevent context pollution, suggesting patterns for delegating tasks to sub-agents. This structured approach aims to optimize performance and maintain clarity in interactions with different AI models.