Developer guide
For developers using the self-service developer portal: get a personal key, call the gateway, see which models you can use, and ask for more. Want the concepts first? See the Overview. New to the gateway itself? The Quick Start and the Integration Guide cover endpoints, base URLs and SDK examples.
The portal still uses some older names. A profile is what admins call an access plan, a model tier is a model option, and a blueprint is an app template.
Open the portal
- You only have self-service access: sign in with your work account. You land in the portal.
- You also use the admin console: open the account menu at the bottom left of the sidebar and select Switch to Self Service. Admin console at the top right of the portal takes you back.
The portal's tabs are My usage, LLM Gateway, MCP Gateway, Workflow agents and Skills. Your access plan can hide some of them.
My usage shows activity attributed to you across your approved apps. Shared-key traffic without a user identity is not included.
Your apps
LLM Gateway lists the apps you can open, with your requests, estimated cost and API keys. The buttons at the top open Analytics, Platform providers, Models and Requests.
Each app card shows its providers and models and what you can do there:
If an app you expect is missing, you do not have viewer access to it. Ask your admin, or request access.
If your access plan gives you one starter app instead of an app list, LLM Gateway opens My developer key: select Issue my developer key to get a personal key in that app. Its value is shown once.
Get a personal key
On an app with personal keys:
- Select Keys & Connect on the app card.
- Under API credentials, select Create API key.
- Enter a Key name that says where it will live, for example
laptoporci-runner. - Choose when it Expires. If the app or your access plan caps key lifetime, the key must expire within that cap.
- Copy the key. It is shown only once.
New keys look like sk-quilr-ss- followed by 48 hex characters. Keys created before November 2026 look like sk-quilr-<app key>:ss1.<...> and keep working.
If the app does not let you view key values, the portal creates the key but does not show its value. Ask your admin for View and copy API key values on that app.
Your keys are listed with their status, expiry and last use. Rename, Edit expiry and Revoke act on one key. You can hold up to 10 active keys per app; your access plan can set a lower limit. One key per machine or environment is a good habit, so you can revoke one without breaking the others.
On an app with shared keys, Keys & Connect shows the app's keys instead. Copy one if the app lets you view key values.
Use the key
The key is an ordinary gateway credential. Gateway quickstart on the same tab shows the gateway URL and examples.
from openai import OpenAI
client = OpenAI(
base_url='https://guardrails-usa-2.quilr.ai/openai_compatible/',
api_key='sk-quilr-ss-...' # your personal key
)
response = client.chat.completions.create(
model='gpt-4o-mini',
messages=[{'role': 'user', 'content': 'Hello!'}]
)
Every request made with your key is attributed to you, so your activity, usage and findings are your own.
Models you can call
Models (on the LLM Gateway page) shows the model tiers your access plan allows. Send a tier's Model name to call as the request's model; the gateway routes it.
- No model tiers apply to you: call your apps' models as usual.
- Only these model names work with your personal keys: other models are refused with
self_service_model_not_allowed, and responses do not name the model that served them. - Otherwise, other tiers are refused, and other models stay available under your organization's policies.
- A reasoning effort cap lowers requests that ask for more. It never refuses them.
These limits apply to your personal keys only. Request a stronger tier asks for more.
Ask for more
Open Requests (on the LLM Gateway page). My requests lists what you asked for; New request asks for something new.
Say why you need it and until when. Every approval ends on a date, at most a year away, except a new application. You cannot approve your own request, and you can withdraw a request while it waits. You can have up to 20 pending requests.
If you own an app, a To review tab lists requests about it for you to approve or reject.
Create your own app
If your access plan lets you create apps, select New application on the LLM Gateway page, pick a Blueprint and name the app. You become its owner, and it starts with the blueprint's providers, routing, settings and guardrail policies. Who can do what, shared keys, budgets, the prompt store, JWT sign-in and key expiry start empty.
On your own app:
- Your access plan decides whether you change its settings directly, ask for changes, or do not see settings.
- You can view its key values and all of its logs.
- You answer requests about it, such as other people asking for access.
- You can disable or delete it.
Past your plan's limit, the button reads Request another application and an administrator decides.
Request a settings change
If Changes need approval shows on the app, open Settings and edit it: your edits are submitted as a request, and nothing is live until an administrator approves it. With direct settings update, your saves apply immediately and are versioned in the app's config history, where an administrator can roll them back.
Track your submissions in the app's Change requests:
Self-service access itself (credential mode and who can do what) is admin-only and is not in the settings editor.
Troubleshooting
Related
- Overview - concepts and the key format.
- Admin guide - how access is set up on the admin side.
- Audit Log - how settings change requests are reviewed and rolled back.
- Integration Guide - endpoint URLs and code examples per SDK.
- Conversation Grouping and Agent Monitoring - make your logs more useful.