Manage a Client's Model Access
This guide explains how a platform administrator controls which models a client uses on a PaletteAI Inference Launchpad appliance. Model access depends on whether a model runs locally on the appliance or is reached through an external inference endpoint.
- Local models. Every client with a valid API token can call every model served locally on the appliance. You do not grant access to local models. To choose which local model handles a client's requests, route the client with a tier map.
- External models. A new client cannot reach external models until external inferencing is turned on for it. External access is denied by default. A client gets it from its workspace policy, or from its own settings when it overrides that policy. Every external model comes from an external inference endpoint that a platform administrator set up under Integrations. Refer to Register an External Inference Endpoint.
To understand how clients and models relate, refer to Clients and Quotas.
Prerequisites
- A running PaletteAI Inference Launchpad appliance, with the console reachable.
- (Tier map only) An existing client. To create one, refer to Create a Client.
- Console access with permission to manage clients. Managing clients can require operator access.
- (External models only) The endpoint set up under Integrations by a platform administrator. Refer to Register an External Inference Endpoint.
Route a Client to Specific Models
Use a client's tier map to route the client's model aliases to the models you choose.
-
In the workspace drop-down menu in the console header, select the workspace that holds the client.
-
From the left main menu, select Access & Policy. The Clients list opens.
-
Select the client to open its detail panel.
-
Select the Routing section.
-
In the Tier map, select Add alias rule, or edit an existing row.
-
Set the Alias prefix, such as
claude-opus-, and the Model it routes to. Local models and external models appear in the picker. An external model is listed as<endpoint> / <model>, such asanthropic / claude-opus-4-8, with a marker that shows the request leaves the appliance. The Tier map starts with seven seeded alias prefixes,claude-fable-,claude-opus-,claude-sonnet-,claude-haiku-,gpt-,gemini-, andgrok-, which you can map but not remove. -
Save the client.
The tier map applies to the selected client. Requests from other clients follow their own tier maps, so they are not
routed to that model unless you configure their tier maps as well. For requests the tier map does not settle, such as a
request that sends auto, refer to Configure Semantic Routing. For how the two
controls combine, refer to Routing Behavior.
Allow a Client to Reach External Models
A client reaches external models through its workspace policy. To let every client in a workspace reach an endpoint, authorize the endpoint on the policy, as described in Authorize a Policy. To give one client its own endpoint settings instead, override the policy when you create the client.
-
In the workspace drop-down menu in the console header, select the workspace for the new client.
-
From the left main menu, select Access & Policy. The Clients list opens.
-
Select Add client. The Add client wizard opens on the Overview step.
-
On the Overview step, enter a Client name, and then turn on Override policy settings.
-
Go to the Egress step, and turn on Enable external inferencing.
If no endpoint is set up under Integrations, the step shows the message
No endpoints are set up yet.instead of the endpoint table. A platform administrator must add an endpoint first. Refer to Register an External Inference Endpoint. -
Select Add endpoint. A row is added to the endpoint table.
-
In the Endpoint column, select an endpoint. The list offers only the endpoints that are set up under Integrations.
-
In the Key column, decide which key the client uses for the endpoint.
The Key field reads What to do Using the Integrations key. Paste one to overrideLeave the field empty to use the key stored under Integrations. To use a different key for this client, paste it. The pasted key applies to this client only. paste tokenPaste a key. The appliance holds no key for this endpoint, so a key is required. The client does not need a key of its own when its policy or Integrations already holds one.
-
Set a positive Daily limit. Egress is fail-closed on the cap, so a limit of
$0blocks that endpoint for the client. -
Repeat steps 6 through 9 for each endpoint the client can use.
-
Complete the remaining steps of the wizard, and then select Create client.
Recommended Setup for Enterprise Anthropic
For the common shape (an enterprise Anthropic subscription, mostly local traffic, occasional flagship), use the following opinionated defaults after you authorize the Anthropic endpoint on a policy, as described in Authorize a Policy, or on a client that overrides its policy, as described in Allow a Client to Reach External Models. For the mental model behind each choice, refer to External Inferencing.
- Have a platform administrator set up the Anthropic endpoint under Integrations with the enterprise key, and leave the Key field empty on every policy and client that should have external access (Pattern A: one contract, one key).
- Set the tier map so
claude-opus-*routes to ananthropic / <model>target andclaude-sonnet-*andclaude-haiku-*route to a local model. This captures most of the intended traffic mix because Claude Code defaults to Sonnet for cheap-tier turns and reserves Opus for hard work. - Set the Daily limit to the monthly cap you agreed with finance, divided by the days in the billing period.
- Configure
on_exceed=localso the client falls back to a local model when the budget runs out, instead of returning HTTP 429. - Leave Enable external inferencing on for the policies and clients that need external access, and off for every other one.
- Skip bursting for now. Enable it later only if local capacity becomes the recurring bottleneck.
To verify the setup, send a request that resolves to the flagship tier through this client and check the response
headers for X-AIBox-Frontier: anthropic on a successful external turn, or X-AIBox-Frontier: budget-exhausted after
the budget runs out. If neither header appears, the client is bypassing the appliance. Refer to
Common Misconfigurations.