← Back to work

Self-Hosted Private AI Gateway

code

A one-compose, self-hosted kit that turns your own Claude, Codex and Gemini subscriptions into an OpenAI-compatible API with a multi-model chat UI.

Self-Hosted Private AI Gateway

A self-hosted kit that turns the AI subscriptions you already pay for — Claude, Codex (GPT) and Gemini — into one private, OpenAI-compatible gateway with a multi-model chat UI. A single docker compose starts the whole stack, on your own machine and with your own credentials.

It's for people with an always-on box — a NAS, a VPS or a side router — who want every model side by side in one place, from a browser, a phone, or any OpenAI-compatible client.

How it fits together

What it is not: it ships no credentials and isn't a reseller of anyone else's quota — you connect your own subscriptions. The README is upfront that proxying subscription credentials may break a provider's terms, so it's meant for personal, self-hosted use. Start, stop and status scripts, a deployment and troubleshooting guide and a credentials guide are included; the kit's own code is MIT-licensed.