TangleBrain

Most AI tooling sends every request to a paid cloud API by default — even when you already run capable models on hardware you own. TangleBrain keeps your whole roster of backends in one editable YAML file and favors the credentials you already hold: local models and OAuth-logged-in tools come first, while raw API keys stay a separate, explicitly-gated opt-in (it never injects a key into a CLI). An optional classifier routes by complexity — grunt work goes to your free local model — and every routed task is logged with an estimated cloud-equivalent cost, so you can see what you're spending versus avoiding. Runs standalone or drops in alongside TangleClaw and the wider Tangle family.