Workflows#

These walkthroughs chain together the individual tasks described in Core features and Administrator features into common, end-to-end scenarios. Each step links back to the page where it is performed.

Connect an AI provider and confirm it is available#

For administrators.

  1. Go to Providers → Add Provider. Select the provider type, enter its API Key and (if needed) a custom Base URL, and optionally list Allowed Models. Click Save.

  2. Make sure the provider is enabled using the toggle switch on its card.

  3. Open the Models page and search for a model from that provider. Use the Provider filter to confirm the model now appears in the list.

  4. As requests start coming in, use the Dashboard (the By Provider table) and the Usage page to confirm activity is routing through the new provider.

Onboard a user with limited access#

For administrators, then the user.

  1. (Admin) Go to Users → Add Restriction. Enter the user’s email or username, specify the Allowed Providers and/or Allowed Models they should have, and click Save.

  2. (User) Log in, go to API Keys → Create Key, give it a descriptive name, and copy the key immediately — it is only shown once.

  3. (User) Open the Models page. Only the models and providers made available to you appear here. Click the copy icon next to a model to copy its identifier, and paste the key and identifier into your script, tool, or IDE plugin.

  4. (User or admin) Confirm the requests are flowing on the Usage page — the user sees their own history, and an admin sees all users’ history from the admin Usage view.

Turn on and verify a guardrail#

For administrators.

  1. Go to Guardrails and enable the guardrail you want (Content Filter, Prompt Injection, or Custom) using its toggle. Changes take effect immediately for all new requests.

  2. Open the Test Guardrails tab, select the guardrail(s), paste sample text, and click Run Test to see whether it triggers and what action it takes (block / mask / warn).

  3. Review Guardrail Logs to confirm the guardrail is catching what it should — filter by guardrail name or trigger status, and check the pass rate to make sure it is not over-blocking legitimate requests.

  4. For any individual request, open its Request Detail panel from the Usage page to see the guardrail events that request triggered.