New API

New API is a self-hosted AI gateway for developers and teams sharing access to multiple model providers. It lets apps and agents keep one connection while administrators change upstream services.

• API compatibility: Convert requests and responses between OpenAI Chat Completions, Responses, Anthropic Messages and Gemini formats, with providers including DeepSeek, Qwen, Azure OpenAI, AWS Bedrock and Vertex AI.
• Provider routing: Distribute requests using priorities and weights, map model names and retry failed requests across connected services.
• Team access: Control permissions for users, groups and API keys, with OAuth/OIDC, passkeys and two-factor authentication for sign-in.
• Usage and costs: Review usage, cost accounting and audit logs in the web console, and try models in its playground.
• Beyond chat: Connect image, audio, embedding and reranking APIs, or use JavaScript plugins for video and other asynchronous tasks.

The gateway runs locally or on your server through Docker and is open source under AGPL-3.0. Model requests go to the connected upstream services. Cloud providers require internet access and authorized API keys or accounts; those providers receive the requests even when you host the gateway yourself.

Streaming, tool use, reasoning and multimodal inputs depend on the upstream model and conversion path. Some protocol-specific fields don't translate exactly, and WebSocket support requires a compatible upstream and channel configuration.