New API

New API solves a real problem: every AI provider has a different API format, different pricing, and different authentication. Instead of writing integration code for each one, you point your application at New API and it translates everything into a single OpenAI-compatible format. The project is a fork of One API, rebuilt with a React frontend and more features. It supports over 40 AI providers including OpenAI, Anthropic, Google Gemini, Mistral, Cohere, DeepSeek, and local models through Ollama. Every provider gets mapped to the same chat completion, embedding, and image generation endpoints, so your application code stays the same regardless of which model you call. Key management is the core feature. You add API keys from multiple providers, and New API handles routing, rate limiting, and failover automatically. If one provider goes down, traffic shifts to the next available key. If a key hits its rate limit, the next key picks up the request. You get a single dashboard showing usage across all providers, per-key spend tracking, and token-level billing. The billing system supports per-user quotas and pay-per-token models. You can resell AI access to your own users with markup, or just track costs internally. Each user gets their own API key that routes through your provider keys, with spend limits and model access controls. New API runs as a Docker container or a single binary. It uses SQLite by default and supports MySQL and PostgreSQL for production deployments. The admin panel handles everything: adding providers, managing users, viewing logs, and configuring channel priorities. For teams that use multiple AI models, New API eliminates the integration overhead. One endpoint, one authentication system, one billing pipeline. Your code calls /v1/chat/completions and New API figures out the rest.