Quickstart

Get your first request routed through Routero AI in under 5 minutes. If you already use the OpenAI SDK, this is a one-line change.

Prerequisites: An API key from platform.routero.ai. If you don’t have one yet, sign up free — no credit card required.

Three ways to integrate

Approach Best for What changes
OpenAI SDK (drop-in) Existing OpenAI codebases base_url only
Direct REST API Any language, no SDK dependency —
litellm Python library In-process routing, no proxy server model prefix

Option 1 — OpenAI SDK drop-in (recommended)

Change base_url to https://api.routero.ai/v1. Everything else — messages, tools, streaming, vision, structured outputs — stays identical.

Python 

import openai

client = openai.OpenAI(
    api_key="YOUR_ROUTERO_KEY",          # your Routero virtual key
    base_url="https://api.routero.ai/v1",
)

response = client.chat.completions.create(
    model="smart/balanced",              # smart alias, or "openai/gpt-4o", "anthropic/claude-sonnet-4-6", etc.
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

TypeScript / Node 

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "YOUR_ROUTERO_KEY",
  baseURL: "https://api.routero.ai/v1",
});

const response = await client.chat.completions.create({
  model: "smart/balanced",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0].message.content);

curl 

curl https://api.routero.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_ROUTERO_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "smart/balanced",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Option 2 — Direct REST API

The gateway exposes a standard OpenAI-compatible REST interface. Use any HTTP client.

Base URL: https://api.routero.ai/v1

Authentication: Authorization: Bearer YOUR_ROUTERO_KEY

Supported endpoints: /chat/completions · /completions · /embeddings · /images/generations · /audio/speech · /audio/transcriptions · /rerank · /batches · /models and more. See Unified API for the full list.

Model strings

Routero passes any model string to the appropriate provider. You can use:

Format Example What it does
Smart alias smart/balanced Routes to the best available model for the task
Provider-scoped openai/gpt-4o Pins to a specific provider
Bare model name gpt-4o Routero infers the provider
Provider variant bedrock/anthropic.claude-sonnet-4-6 Fully-qualified AWS Bedrock model

Smart aliases (smart/balanced, smart/fast, smart/cheap) are configured in your workspace’s routing policy and apply fallback chains automatically. See Policy Routing.

What just happened

Every request you sent ran through Routero’s four-decision pipeline:

  1. Policy gate — Routero checked your key’s identity, evaluated your workspace’s routing policy (content checks, model allowlist, budget state), and decided whether to proceed.
  2. Provider selection — The Router scored eligible deployments by current health, latency, cost, and data residency. smart/balanced resolved to your configured primary provider.
  3. Account & audit — The token count and cost were calculated and debited from your workspace budget atomically. The full decision was written to your immutable audit log within milliseconds.
  4. Response — The provider’s response was streamed back through the gateway with zero buffering.

The request is now in your platform dashboard under Audit Log.

Activate Advanced Features

Pass any combination of feature IDs on the same request to unlock Routero’s production-AI layer. The proxy resolves each config from your workspace, applies it as a hook, and strips the ID before the upstream call — your application code never changes.

response = client.chat.completions.create(
    model="smart/balanced",
    messages=[{"role": "user", "content": "Summarise last quarter's results."}],
    extra_body={
        "guardrail_id":        "my-pii-guardrail",      # redact PII before the model sees it
        "token_saving_plan_id": "my-cache-plan",         # compress + cache the response
        "prompt_id":           "my-analyst-system-prompt", # inject versioned system prompt
        "memory_id":           "user-alice-session",    # retrieve Alice's long-term context
    },
)

→ Advanced Features

What to do next