Skip to content
Collab with Anthropic and OpenAI
NovaRouter

One Gateway.
Unlimited Intelligence.

Route every request through a single OpenAI-compatible endpoint. Six frontier models across four providers, automatic failover, and no charge to use it.

Start building freeRead the docs
  • Drop-in OpenAI SDK
  • Streaming on every model
  • Automatic failover
  • Completely free
Compatible routing network
Collab with Anthropic + OpenAI
Anthropic
OpenAI
Claude
OpenAI
NovaRouter
One Gateway. Infinite Possibilities.
route.tsgateway live
import OpenAI from "openai";const nova = new OpenAI({  baseURL: "https://nova-router-main.vercel.app/api/v1",  apiKey: process.env.NOVAROUTER_API_KEY,});const res = await nova.chat.completions.create({  model: "claude-sonnet-5",     // any of the six aliases  messages: [{ role: "user", content: "Ship it." }],});
Routing decision
  1. Request receivedus-east-1
  2. Scoring 6 candidates6ms
  3. Selected Claude-Sonnet-5best cost/quality
  4. First token streamedstreaming
  • 9
    Models available
    8 chat · 1 embedding
  • 6
    Publishers routed
    Automatic failover between upstreams
  • 2M
    Largest context
    Tokens, on Gemini-3 Pro
  • 4
    Endpoints to learn
    models · chat/completions · responses · embeddings
The network

Compatible With Leading AI Providers

One OpenAI-compatible endpoint in front of 9 models from 6 providers. Switch between them by changing a single string — no new SDK, no rewritten call sites, no per-provider billing.

Collab with Anthropic and OpenAI
Anthropic3 models
Z.ai2 models
OpenAI1 model
Google1 model
DeepSeek1 model
xAI1 model

Provider names are trademarks of their respective owners and are shown to describe compatibility. NovaRouter is not affiliated with or endorsed by them.

Everything in one place

Built to be swapped into, not migrated to

Change the base URL and the model string. Everything else — SDK, call sites, error handling — stays exactly as it is.

Catalogue

9 models, one schema

Compare speed, context, streaming support and price side by side before you commit to one.

Browse the catalogue
Routing

Fallbacks without retry code

Declare an ordered chain. Rate limits and 5xx responses reroute to the next model automatically.

Read the docs
Pricing

Completely free

Every model, unlimited requests, no plan and no card. Requests are logged so you can see your own usage — nothing is billed.

See pricing

Highest quality right now

All 9 models
  • Anthropic logoFable-5Anthropic1M ctx128K out$10/1M in
  • Anthropic logoClaude-Opus-5Anthropic1M ctx128K out$15/1M in
  • OpenAI logoGPT-5.6 SolOpenAI400K ctx128K out$5/1M in
The platform

One integration, every frontier model

Pricing, context limits, model kind and streaming support are visible before you commit — and switching between models is a one-line change.

9
Models
Frontier and open-weight, one schema
6
Providers
Anthropic · OpenAI · Google · DeepSeek · xAI · Z.ai
2M
Largest context
Tokens, on Gemini-3 Pro
8/9
Stream tokens
Server-sent events on every request
$0.02
Lowest input cost
Per 1M tokens, on Nova-Embed-1
1
Endpoint to learn
OpenAI-compatible — change the base URL
Why NovaRouter

9 models, 6 providers, one integration

Swap models without touching product code, survive a provider incident through ordered failover, and inspect usage from one dashboard.

One endpoint, every model

Point the OpenAI SDK at our base URL and change the model string. No new client, no rewritten call sites.

Automatic failover

Declare an ordered fallback chain. Rate limits and 5xx responses reroute to the next model without a retry in your code.

No bill at all

Every model in the catalogue, unlimited requests, no plan to choose and no card to add. Per-request logs so you can still see exactly what you used.

Start routing in the next five minutes

Create an account, confirm your address, and every model in the catalogue is yours to call. No card, no plan, no charge.

Continue with GoogleRead the docs