Calling the API

Routero exposes a fully OpenAI-compatible API. Point any OpenAI SDK — or a plain HTTP client — at Routero with two values, and everything else (messages, tools, streaming, vision, structured outputs) works unchanged.

• Base URL: https://api.routero.dcsmartvision.com/v1
• Authentication: Authorization: Bearer YOUR_ROUTERO_KEY (a Routero virtual key)

Create a virtual key in the dashboard under API Keys. You can scope a key to a team, limit its models, and attach a budget. Treat it like a secret — it authenticates every request your application makes.

---

Your first request (curl)

curl https://api.routero.dcsmartvision.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_ROUTERO_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-5.5",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

---

Python (openai)

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_ROUTERO_KEY",
    base_url="https://api.routero.dcsmartvision.com/v1",
)

response = client.chat.completions.create(
    model="openai/gpt-5.5",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)

---

TypeScript / Node (openai)

import OpenAI from "openai";

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

const response = await client.chat.completions.create({
  model: "openai/gpt-5.5",
  messages: [{ role: "user", content: "Hello!" }],
});

---

Models

The model field accepts any model string Routero is configured to serve. See Models for the model list and Routing & Load Balancing for how routing works.

---

Streaming

Set stream: true to receive tokens as they are generated — identical to the OpenAI streaming format:

stream = client.chat.completions.create(
    model="openai/gpt-5.5",
    messages=[{"role": "user", "content": "Hello!"}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

---

AI capabilities on a request

Pass the AI capability IDs (guardrails, prompts, memory, token saving) through the SDK’s extra_body, or as extra JSON fields in the request body. Routero applies them inside the gateway and strips them before the upstream provider sees the request.

response = client.chat.completions.create(
    model="openai/gpt-5.5",
    messages=[...],
    extra_body={"guardrail_id": "pii-redact-prod"},
)

---

Next

→ Cursor and Claude Code to route your coding assistants through Routero. → Inference APIs for the full endpoint surface.