NewAPI

NewAPI is an open-source AI API gateway for connecting multiple model providers through a single OpenAI-compatible interface. According to the vendor, it can connect more than 40 upstream providers—including OpenAI, Anthropic Claude, Google Gemini, DeepSeek, xAI Grok, Azure OpenAI, AWS Bedrock, Moonshot, Qwen, GLM, and others—while centralizing channel routing, API keys, quotas, and usage logs in a self-hosted control panel. See the official NewAPI site for the project and deployment entry point.

What it does

NewAPI standardizes access to connected providers behind one API endpoint that follows OpenAI request and response formats. The vendor says this allows applications using OpenAI-compatible SDKs to switch models or providers by changing the base URL and using a NewAPI-issued key, without other code changes.

Its channel-based routing maps model names to upstream providers and credentials. Administrators can create multiple channels for the same model, then configure load-balancing weights, priority ordering, and automatic failover across channels. This provides a single routing layer for applications that need to distribute requests or reroute them when a channel is unavailable.

Who it helps

NewAPI is aimed at teams operating internal AI tools, public-facing applications, or multi-provider workflows. It can provide a shared gateway for conversational, generative, and multimodal requests while keeping provider credentials out of end-user applications.

It is also relevant for teams that want to retain an OpenAI-compatible integration while gaining centralized control over provider access, user groups, quotas, and usage visibility. The product is self-hosted and open source, so the vendor positions it for organizations that want data and gateway operations to remain on their own infrastructure.

Workflow

A typical setup starts by deploying NewAPI and adding upstream providers and model routes. Administrators then configure channels, map models to providers and keys, and issue separate NewAPI credentials to users or applications.

Issued keys can have individual quota limits, rate limits, and group-based access policies. The vendor states that upstream provider credentials remain stored on the server and are not exposed to end users. Each request is recorded with details such as model, token count, latency, and status; the management panel provides per-user and per-channel consumption views, request logs, and usage or billing visibility.

For deployment, the project supports Docker or a standalone binary and lists compatibility with SQLite, MySQL, and PostgreSQL. The supplied documentation describes Linux-server deployment but does not specify hosted plans or pricing.

Strengths and limits

NewAPI’s documented strengths are its broad provider aggregation, OpenAI-compatible access, configurable routing and failover, separated user credentials, and self-hosted operational control. It can reduce the need to build separate gateway and quota-management workflows for each provider.

The available evidence does not establish provider uptime, routing performance, security audit results, exact support coverage for every listed model or API feature, or commercial support terms. Teams should therefore verify compatibility, operational requirements, and the current provider list in the official documentation and site before deployment.