Quick Start
Create an app, configure your SDK to use the gateway, and send a first request.
Before you start
You need one of:
- Credentials for a provider (for example an OpenAI API key), or
- A platform provider already set up in Settings > AI Gateway > Models (V2 console only, see Providers and Models). You can also add one from inside Create App.
1. Create the app
Go to Settings > AI Gateway > LLM Gateway and click Create App. The wizard has two steps.
Step 1: Application and providers
Next, under Where should this app get its models?, choose where the provider credentials come from.
The Platform providers option exists only in the V2 console, and it is the default. In V1, every app keeps its own provider credentials. Either way, the choice is fixed once the app is created.
Platform providers. Under Available, tick one or more providers. The first one you tick is the Primary; use the up and down arrows to change the order. A disabled provider is skipped, and when the primary fails the gateway fails over to the next provider in the order (see When a provider fails). The app inherits all their models and credentials, including models added later. If none exist yet, click Add a platform provider to add one without leaving the form, or Use app-only credentials.
App-only credentials. Select App-only credentials and select a provider tile. For providers with more than one API, also select one under Which API does the gateway talk to?.
- Keep or change the Provider label.
- Fill in the credential fields. Each provider's fields are listed in Provider Support.
- Under Models, click Get available models and choose from the list, or type a model in Add a model by name and click Add. Each model is checked against the provider before it is added.
- Optional: click Add another provider to add more providers for routing.
To connect an app to QuilrAI-provided models, add a Custom endpoint provider with the Chat completions API (general), base URL https://models.quilrai.dev/v1 and a model API key from Settings > AI Gateway > Models > API keys.
Click Create app. There is no guardrail step: every new app starts with these defaults, which you can change later in its Security Guardrails section.
Monitor logs detections without changing traffic.
Step 2: Key details
The last step, Integration, shows the new Quilr key, the app's log export key and, if you have access, the all-apps log key. Copy them now, then click Done. You can copy the Quilr key again later from the app's API Integration section (see Applications and Keys).
2. Send your first request
Point your client at a regional gateway URL and use the Quilr key. Nothing else changes.
from openai import OpenAI
client = OpenAI(
base_url='https://guardrails-usa-2.quilr.ai/openai_compatible/',
api_key='sk-quilr-xxx', # your Quilr key
)
resp = client.chat.completions.create(
model='gpt-4.1-mini', # a model enabled on the app
messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(resp.choices[0].message.content)
The example uses US East. Select the nearest region and the appropriate path for your SDK (Anthropic, boto3, Vertex AI, Responses, Realtime), in the Integration Guide.
3. Check that it worked
On the app card, click Logs. The Activity > Requests view lists each call with its model, provider, guardrail outcome, HTTP status, tokens and latency. See Overview for the other tabs.
4. Tune the app
Open the app card's Configure buttons to change any of these settings later.
If the Policy Engine is on, guardrails, Guardian Agent, limits, token saving, routing, identity requirements and Prompt Store enforcement follow published policies instead of these app settings.