quantumnous/new-api

New API is a self-hosted AI gateway and management system written in Go that acts as a central hub for multiple large language model providers. The problem it solves is practical and common: teams or individuals use several AI services, OpenAI, Anthropic Claude, Google Gemini, DeepSeek, and others, each with its own API format, authentication scheme, and billing system. Juggling multiple API keys, usage quotas, and incompatible request formats across different applications is messy. New API sits in front of all those providers and presents a single unified endpoint, converting incoming requests between OpenAI-compatible, Claude-compatible, and Gemini-compatible formats automatically. The gateway works by accepting API calls at a single address, routing them to the appropriate backend provider based on configuration, and translating the request and response formats as needed. An admin dashboard lets you manage user accounts, assign API keys to groups, set per-model quotas, restrict which models specific tokens may use, and view usage statistics. It also handles billing, you can charge users per token consumed, set up online recharge via Stripe or EPay, and configure cache billing to avoid double-charging for repeated identical requests. Multiple authentication methods are supported, including Discord, Telegram, and OIDC single sign-on. You would use New API when running an internal AI service for an organization where you want to centralize model access behind a single endpoint, control costs per user or team, or resell AI API credits. Personal users also use it to consolidate multiple provider keys into one convenient local proxy. The tech stack is Go for the backend, with a web frontend. It is deployed via Docker and supports SQLite or MySQL for storage.