Skip to main content
You can run Millwork tasks through your own provider account and have the provider bill model usage. Provider setup lists the nine offered providers, their commands and the access to prepare. For OpenRouter, you sign in and approve access in the browser; you do not need to find or paste a provider key. Millwork tests the connection and adds its available models to your catalog, the list of models your organization can use. Your app sends tasks with its Millwork API key. A provider connection is the saved link between your organization and the provider account. Check GET /v1/model-source-profiles for the providers your organization can connect. This page explains what you pay and how the connection works. The task itself, with every request and its expected result, is Connect a provider account.

​
What you pay for what

Millwork charges nothing for the model usage itself on runs through your provider account. Only the platform fee applies, and a run you cancel keeps it. A test run (the API calls it Echo, mode: "echo") does not call a model, so it has no platform fee and no model cost. Read Model access and billing for the full rule and the one case where Millwork returns the fee. Millwork never returns your provider credentials. They do not appear in an API response, a result, a receipt, or a support report. Revoking the connection in Millwork stops later runs from using it. It does not revoke the key at the provider. Manage that access in the provider’s own controls. A short-term Bedrock key or STS session keeps its existing AWS expiration.

​
Is this path for me?

Use your provider account when your organization already has an account with an AI provider or cloud, and you want that provider to keep billing you for model usage. If you have no provider account, ask support whether a model Millwork provides is available to your organization.

​
How the connection works

For OpenRouter, the CLI waits while you approve in the browser, then checks the connection and loads its models:
1

Choose a provider

GET /v1/model-source-profiles lists the providers and sign-in methods your organization can use. You can connect only a provider and sign-in method from that list.
2

Complete secure setup

For OpenRouter, follow CLI setup: sign in and approve access in the browser, then return to the same terminal. For a key-entry provider, enter its key on Millwork’s private setup page. For AWS Bedrock, enter a short-term Bedrock API key there, or choose the advanced STS method explicitly. Return to the same CLI journey, or continue the API connection steps if that is how you started. Keep credentials and browser links out of chat.
3

Test the connection

Millwork checks whether the credential can access the provider. A connection that has not passed its test cannot serve a run. If the test fails, Provider connection help explains each test error.
4

Refresh your models

Millwork asks the provider which supported models the account can use and adds them to your catalog, billed to your provider account. An empty catalog after a passing test needs a check of the account’s model access and the models Millwork supports. It does not by itself mean the credential is invalid.

Connect a provider account

The complete request sequence: every command, its expected result, and the id to keep for the next step.

​
After you connect

After a completed live run on one of the new catalog models, read its answer and matching receipt. The receipt records the selected provider and model; provider-reported model details are separate and may be absent. Millwork may pick a different saved model when several fit the task; read the receipt’s route to see which one was selected and who bills its usage. Pinning a run to one of these saved models also keeps it on your provider account. To pin it, send the model’s arm_id in routing.required_arm_id. The run then uses that model and no other. If that model is not ready or cannot take the task, Millwork refuses the request with invalid_state and charges nothing.

​
AWS Bedrock

Choose AWS Bedrock when it appears in millwork provider list. AWS bills model usage to your account; Millwork charges its platform fee when it accepts a live run. Connect with a short-term Bedrock API key; Millwork lists your system profiles and matches the selected model. The source Region defaults to us-east-1 and must match where you generated the key. AWS STS credentials are an advanced option. Follow AWS Bedrock setup from the private credential form through the model’s answer and receipt. The guide also shows how to renew access on the same connection without changing its method. Azure Foundry and Vertex AI are not available today. Gemini means the Gemini Developer API, not Vertex AI. If your provider is listed but setup fails, email support. Do not include the key or the setup link.

​
Next step

Run a model

Save a catalog model, submit a live task, and read which model and provider answered.
To renew or replace access, follow Replace a provider key. Millwork tests the replacement before switching and keeps the previous settings if the test fails. To stop using the account, disconnect it. Calls already sent may finish; removal of Millwork’s saved credential is scheduled separately from provider-side revocation.