Routin AI Gateway

Turn model trafficinto one lightpath.

A production AI API gateway that brings OpenAI-compatible requests, provider routing, metering, organization access, and developer tools into one stable path.

Open docsAPI reference
Routin AIrouting online
ProtocolOpenAI compatible
ProvidersMulti route
ControlQuota + org
01Ingress

Keep your SDKs, enter the gateway

Point existing OpenAI-compatible clients at Routin and keep request shape, streaming, and tool flows familiar.

Read Chat Completions
02Routing

Distribute by model, plan, and availability

Organize traffic across OpenAI, Anthropic, Gemini, and other providers without binding the app to one upstream.

Read Responses
03Metering

Make usage and quota visible

Plans, orders, quotas, snapshots, and billing boundaries move through one policy surface.

Read billing
04Access

Separate organizations, keys, and rules

Split credentials and permissions by team, app, and member so production access is clear and revocable.

Read organization
05Tools

Connect developer tools on the same path

Codex, Claude Code, Gemini, Cherry Studio, Kilo Code, and OpenCode can all use the same gateway settings.

Read tool setup
baseURL: /v1
model: auto
stream: true
OpenAIAnthropicGeminiAzureVolcengine
Quota live
OrgKeysPlans
CodexClaude CodeGemini CLICherry Studio
API

AI gateway architecture

Three-plane gateway architecture

Routin separates each model call into clear layers: clients enter through the compatible protocol surface, policy decides quota and access, routing selects providers, and observability writes usage, latency, and cost back into the business system.

01 ClientApps and tools

Web apps, servers, Codex, Claude Code, and other clients enter through one gateway.

02 ProtocolOpenAI-compatible layer

Unifies Chat, Responses, embeddings, images, audio, video, and related request shapes.

03 PolicyPolicy and access

Applies quota, access, and audit rules by organization, plan, key, and member identity.

04 RouterModel router

Chooses upstream providers by model, provider health, plan capability, and failover policy.

05 ProvidersProvider adapters

Normalizes provider responses, errors, and capability differences into one exit surface.

06 BillingMetering and billing

Records tokens, media jobs, plan windows, and order state into explainable cost data.

07 ObservabilityTelemetry and alerts

Tracks success rate, latency, errors, provider state, and request traces continuously.

08 SecretsCredential boundary

Keeps user keys, provider keys, and app credentials isolated inside controlled boundaries.

OpenAIAnthropicGeminiAzureVolcengine

Workflows

Start where your migration is happening

The docs follow production rollout order: connect requests, then shape routing, metering, access, and tools.

Step 01API access

Chat Completions, Responses, images, audio, video, and embeddings.

Step 02Model routing

One model entry point with provider configuration and migration notes.

Step 03Plan metering

Subscriptions, quotas, usage snapshots, and cost boundaries.

Step 04Developer tools

Connect clients and command-line tools to the same gateway path.

Quick start

Change one base URL, then open more control

Start with compatible access, then enable provider routing, metering policy, organization access, and tool configuration.

Open setup guide
import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.ROUTIN_API_KEY,
  baseURL: 'https://api.routin.ai/v1',
});

const response = await client.chat.completions.create({
  model: 'auto',
  stream: true,
  messages: [{ role: 'user', content: 'Route this request.' }],
});

Get started

Move model requests onto a clear path

Begin with the API closest to your existing app, ship the first request, then expand into team and production control.

Open documentationGitHub