spacepirate15/quantum-free-router

Most AI language model providers offer free tiers with rate limits, daily quotas, and occasional downtime. If you are running an automated task or AI agent that relies on a single free endpoint, hitting any of those limits stops your work. This project, quantum-free-router, sets up a local routing layer that sits between your code and the various free-tier AI providers so that when one fails, another takes over automatically. The router runs as a background service on your computer and exposes a single URL at port 4000 that behaves like an OpenAI-compatible API. You point your code at that local address instead of directly at any one provider. Behind the scenes, the router tries providers in a documented priority order: OpenCode Zen, KiloCode, NVIDIA NIM, Mistral, Cerebras, Groq, SambaNova, Cohere, and Gemini, among others. Each provider in the list has been tested against the router and given a status: active (working and certified), retry-later (temporarily unreliable), or quarantine (excluded until issues are resolved). The project is built on an open-source routing tool called Bifrost. Installation takes one command on Linux or Windows Subsystem for Linux: it downloads and runs an install script that sets everything up as a systemd service. You then add your free API keys to a config file and the router handles the rest. For developers running tools like coding agents that need to stay active across long tasks, the main benefit is continuity. The router does not bypass provider limits or give you more free usage, it gives you a cleaner way to spread requests across multiple free accounts so that hitting one limit does not kill the whole session. All keys stay in a local config file and are never committed to the repository. MIT licensed.