---
title: Can an LLM gateway reuse your existing Okta app for SSO?
url: https://opencontroller.lyzr.com/blog/reuse-existing-okta-app-for-llm-gateway-sso
published: 2026-11-17
updated: 2026-11-17
author: Lyzr
---

# Can an LLM gateway reuse your existing Okta app for SSO?

**Yes, technically: one Okta OIDC app integration can hold several sign-in redirect URIs, so a gateway console can share an app with another governance tool. But a separate app integration per tool is usually the better choice, because each tool then gets its own client secret, group assignments and audit trail. Reuse makes sense only when the tools are one product with one admin population.**

The question usually surfaces mid-rollout. A new gateway build ships with SSO support, the team already has an Okta app for the governance console, and someone asks whether to add a redirect URI or file a ticket for a new app. Here is how to decide and what to configure either way.

## What does Okta allow in a single OIDC app integration?

Okta's OIDC app integration wizard supports three app types: Web Application, Single-Page Application and Native Application. For a gateway console, Web Application is the normal choice. The relevant facts from Okta's documentation:

- **Multiple sign-in redirect URIs.** You can list several absolute URIs in any order. A sign-in that returns to an unregistered URI fails with an error.
- **Wildcards only for subdomains.** There is an option to allow a wildcard in the sign-in redirect URI, limited to subdomains, and Okta states that using it is discouraged.
- **Sign-out redirect URIs.** Also multiple and absolute, with no wildcards.
- **Grant types.** Web apps can use Authorization Code, Refresh Token, Client Credentials and others. PKCE is a client authentication setting, not a grant type.
- **Assignments.** Access can go to everyone, to selected groups, or be left unassigned for now.

So adding the gateway's callback URL to an existing app is a two-minute change. Whether you should is a different question.

## Should you reuse the app or create a new one?

| Factor | Reuse one app | Separate app per tool |
|---|---|---|
| Client secret | Shared; rotating it touches both tools | Independent rotation |
| Who can sign in | Same group assignment for both | Different groups per tool |
| Okta system log | Sign-ins mixed under one app | Clean per-tool reporting |
| Role mapping | Both tools read the same groups claim | Claims tuned per tool |
| Decommissioning | Removing one tool means editing a shared app | Delete the app |
| Setup effort | Lowest | One extra app integration |

Reuse is reasonable when the gateway console and the governance console are parts of the same product, deployed together, used by the same administrators. Otherwise, the separate app wins on blast radius: a leaked secret or an over-broad group assignment affects one tool, not two.

## What do you need to configure for the gateway?

Whichever route you choose, the checklist is the same:

1. **Deploy the SSO-capable build first.** Redirect URIs point at a live callback; configuring Okta before the endpoint exists just produces failed test logins.
2. **Register the exact callback URI.** Use the gateway's documented path on its production hostname. Add a separate URI per environment rather than a wildcard.
3. **Set the sign-out redirect URI** so sessions end cleanly.
4. **Choose Authorization Code** for the console, with PKCE if the gateway supports it. Enable refresh tokens only if the console needs long sessions.
5. **Give the gateway its client ID, client secret and issuer URL.** Store the secret in a secrets manager, not in the deployment manifest.
6. **Assign groups, not everyone.** Limit access to the groups that administer AI traffic.
7. **Map groups to roles.** Send a groups claim and map it to gateway roles such as admin and viewer, as described in [mapping IdP groups to roles](/blog/map-idp-groups-to-roles-in-ai-governance-tools).
8. **Remove local logins.** Once SSO works, disable password accounts except documented break-glass ones; see [SSO-only access for AI governance tools](/blog/sso-only-access-for-ai-governance-tools).
9. **Test the leaver path.** Deactivate a test user in Okta and confirm they lose console access.

## Does console SSO cover agents calling the gateway?

No. SSO governs people signing in to the console. Agents and applications calling the model endpoint need their own credentials: scoped keys per agent, OAuth client credentials, or workload identity federation. Mixing the two is a common first-review finding, covered in [how apps should authenticate to an LLM gateway](/blog/how-should-apps-authenticate-to-an-llm-gateway).

When an agent acts for a specific person, the identity question changes again: you need delegated, on-behalf-of tokens so logs show both the agent and the user. That model is explained in the [agent identity and on-behalf-of delegation guide](/blog/ai-agent-identity-on-behalf-of-delegation).

## How Opencontroller does this

Opencontroller supports OIDC SSO for people and works with open identity standards including OIDC, SAML and SCIM. For agents, it supports delegated on-behalf-of identity and agent workload identity federation, and the Lyzr Gateway attributes cost per agent and per version. It deploys on Lyzr SaaS or customer-controlled infrastructure, with Lyzr engineers embedding for integration work such as IdP setup. See the [capability reference](/capability-reference) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can I use a wildcard redirect URI for gateway environments?**
Okta allows wildcards only for subdomains and discourages them. Register one exact URI per environment instead.

**Does reusing an Okta app share user assignments?**
Yes. Assignments belong to the app integration, so everyone assigned can attempt sign-in to every tool that uses it. Separate apps let you assign different groups.

**Should agents sign in through Okta SSO too?**
No. SSO is for people. Agents need machine credentials or delegated tokens, ideally short-lived and scoped per agent.

## Key facts

- An Okta OIDC app integration (Web, Single-Page or Native) can list multiple absolute sign-in redirect URIs; wildcards are allowed only for subdomains and Okta discourages them; sign-out URIs allow no wildcards.
- Reusing one app integration for an LLM gateway and another governance tool shares the client secret, group assignments and system log; separate apps give independent rotation, assignments and reporting.
- Gateway SSO checklist: deploy the SSO build first, register the exact callback per environment, Authorization Code with PKCE, store the client secret in a secrets manager, assign groups, map groups to roles, remove local logins, test the leaver path.
- Console SSO does not authenticate agents; agents need per-agent machine credentials or delegated on-behalf-of tokens.
- Opencontroller by Lyzr supports OIDC SSO, OIDC/SAML/SCIM standards, delegated on-behalf-of identity and agent workload identity federation.
- Opencontroller pricing starts from $250K per deployment per year.

## Related

- [Should AI governance tools allow local logins or SSO only?](https://opencontroller.lyzr.com/blog/sso-only-access-for-ai-governance-tools.md)
- [How do you map identity provider groups to admin and viewer roles in an AI governance tool?](https://opencontroller.lyzr.com/blog/map-idp-groups-to-roles-in-ai-governance-tools.md)
- [How should users and applications authenticate to an LLM gateway?](https://opencontroller.lyzr.com/blog/how-should-apps-authenticate-to-an-llm-gateway.md)
- [More on Agent identity and access](https://opencontroller.lyzr.com/blog/cluster/agent-identity-and-access)
