click-to-automate/clicktoautomate-ai-nexus-router

ClickToAutomate AI Nexus Router is a free, open source gateway that sits between your applications and multiple AI language model providers, letting you route, manage, and get more value out of requests sent to services like OpenAI, Anthropic, and others. It is built in Go for speed and low resource use, with a React based dashboard for managing it, and ships as either a plain command line executable or a native desktop application built with the Wails framework. Because the whole dashboard is packed into a single Go executable, getting started just means downloading one file and running it, without installing any separate runtime or dependencies. Once running, it automatically switches between different AI providers if one hits a rate limit or fails, so requests keep going through. Any tool that already knows how to talk to an OpenAI style API, including things like Claude Code, Codex, and Cursor, can be pointed at this one local endpoint instead of talking to each provider directly. It also supports sending images along with text prompts to models that can understand pictures, such as GPT-4o and Claude 3.5 Sonnet. There are two ways to run it. The command line version starts a proxy server, by default on port 20128, and prints logs straight to your terminal, though you can still open a web browser to that same port to manage providers even while running this way. The desktop version opens a native window that manages the same Go backend in the background and shows the dashboard directly, with a built in tab for viewing logs. The project is organized into a Go backend that handles the actual API routing and provider management while also serving the built in frontend, a React single page application for the dashboard itself, and a root level Wails wrapper that combines both into one native application. Recent changes added the dual command line and desktop build system, a real time log viewer alongside a background log file writer, smoother auto scrolling and chat deletion in the built in testing playground, proper handling of preflight browser requests so the dashboard can be used directly, and support for sending images to vision capable models. Building the desktop version for Windows uses a single Wails command that produces the finished executable.