Skip to main content
Goal: Make an existing provider account available through Millwork without putting its credential in application requests. You are done when: the connection test passes and its supported models appear in the catalog. Keep your existing provider account and billing relationship. Millwork collects the credential in a separate secure step, tests the connection, and shows only models the connection can use.

​
Is this path for me?

Use this path when:
  • your organization already has an AI provider or cloud account;
  • you want the provider to keep billing your organization; and
  • you want Millwork to run through that account without putting the provider credential in application requests.

​
What your organization owns

​
What this changes

Your provider account and provider bill stay with your organization. Your application sends tasks to Millwork using only its organization API key. Provider credentials are entered and rotated through a separate secure setup.

​
How it works

1

Choose an available provider

Millwork lists the providers and sign-in methods your organization can use.
2

Complete secure setup

Enter the provider credential in a one-time browser step. Do not put it in a Millwork API request.
3

Test the connection

Millwork checks the selected provider account, project, or region. A connection is not ready until the test passes.
4

Refresh your models

Millwork checks which supported models the connection can use. Available models then appear in your catalog.

​
What success looks like

  • the connection test passes;
  • the model refresh completes;
  • supported models appear in your catalog; and
  • later results and receipts identify the provider path that answered when Millwork selects one of those models.
To confirm the provider on a live run, add one of its catalog models and inspect the result and receipt. Millwork may choose another eligible saved model when several match.

Connect a provider

Follow the complete API sequence and check each result.

​
AWS Bedrock

Bedrock uses your provider account only. Before setup, choose a region and an authorized inference profile. The AWS credential must be allowed to read that profile and invoke through it. Millwork stops before inference when it cannot confirm the profile. Azure Foundry and Vertex AI are not available in this release. Gemini means the Gemini Developer API, not Vertex AI. If your provider is listed but the flow fails, email support without sending credentials.

​
Next steps

  1. Run one model from the catalog.
  2. Confirm the provider in the result and receipt.
  3. Add safe polling and retries.
  4. Rotate or revoke the connection when the provider credential changes.