What is one-api?

One API lets you bring all your AI models under one roof. Instead of juggling separate API keys and setup instructions for OpenAI, Anthropic Claude, Google Gemini, and a dozen Chinese AI services, you connect them all to a single hub. You then hand out your own access keys to your team or customers, and everyone can use any of those models through a single, consistent interface. The system acts as a smart middleman between your users and the AI providers. You plug in your original API keys from the various services, and the hub translates requests so they work with whichever model you want to use. It handles load balancing across providers, can automatically retry if a service fails, and tracks how much each user or token is spending. You can set spending limits, create user groups with different pricing rates, and even generate redemption codes for account top-ups. This is built for teams and businesses that rely on multiple AI models and need to manage who can use what. A startup offering an AI-powered product could use it to give their developers access to different models without exposing the original credentials. An agency reselling AI access could use it to create separate keys for each client, track their usage, and control how much they can spend. It also works as a personal gateway if you use several AI tools yourself and want one place to manage them all. The project is notable for being extremely easy to deploy. It comes as a single file or a ready-to-use Docker image, so you can get it running with one command. It supports multi-server setups for handling high traffic, and it offers a customizable web interface where you can add your own branding, set up user login, and configure channels. The code is open source under an MIT license, with the only condition being that you keep a credit link in the footer unless you get permission to remove it.