OmniRoute

OmniRoute is published by the GitHub organization cheaper-inference, with Diego Souza (diegosouzapw) named as maintainer in the README, which calls it "The Free AI Gateway." If you use several AI model providers, it gathers their accounts, API keys and quotas in one place and exposes a single compatible endpoint, so your AI client only needs one address. You can run it as a desktop app, a web app (PWA), a CLI or an MCP server. The README says the interface comes in 67 languages and that it runs on macOS, Windows, Linux, Docker and Android Termux.

Who it's for:developers who juggle several model providers and want one place to manage keys and quotas. Not useful if you don't write code.

Before you start: after a global npm install it runs locally on port 20128, and you point your client at the /v1 address; Docker, Podman and a desktop app are also available. OmniRoute itself is free, but you bring your own upstream accounts and keys, which may cost money. The control panel runs locally and your prompts go to whichever upstream you choose; keys are stored with AES-256-GCM encryption and telemetry is off by default. We don't recommend three features. The first is connecting third-party services through web cookies or an automated browser login, which reuses the sign-in state in your browser. The second is the optional Linux TPROXY transparent proxy. It needs root, changes your firewall rules and installs a local root certificate, then intercepts and decrypts the HTTPS traffic on your computer. The third is JA3/JA4 TLS fingerprint impersonation, which makes the program pose as other software or browsers when it connects. Licensed under MIT. Based on the project's public documentation. Not tested by AI Inspiration World.