new-api
A self-hosted backend for multi-model API aggregation, quota control, billing, and distribution — turn multiple LLM keys into a single managed service.
Tool overview
new-api is an open-source AI API gateway written in Go. It serves as a centralized hub that translates diverse LLM provider APIs into standard OpenAI, Claude, or Gemini formats. Beyond simple proxying, it includes user and token management, multi-channel routing, usage quotas, pricing models, and a billing/charging system — essentially turning raw model APIs into a manageable, business-ready service.
Its core strengths are: unified access to 40+ upstream models, intelligent failover, fine-grained rate limiting per user/key/model, support for prepaid quotas, subscriptions, and pay-per-call billing, and a visual admin panel. This drastically reduces the complexity of building a multi-model AI product from scratch.
On the flip side, it requires self-hosting (recommended 4 core/8 GB RAM, PostgreSQL, Redis) and some operational know-how. Configuration for channel pricing and user plans can be non-trivial. It is overkill if you only need to forward a single provider's API. The codebase is AGPLv3 licensed, which may raise compliance concerns, though a commercial license is available.