> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getmillwork.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Evaluate Millwork

> Check the current value and limits of the private preview.

Use this evaluation to answer three questions:

1. Can my application run a task through Millwork?
2. Can I set limits and confirm what answered?
3. Does the private preview fit my data and operating requirements?

## Before you begin

* Your Millwork organization is approved.
* You created an organization API key.
* You will use public information or test data created for this evaluation only.
* Your catalog contains a model, or you may connect a provider account.
* You do not require an SLA, certification, residency commitment, or published
  pricing for model access provided by Millwork for this evaluation.

## Evaluation plan

### 1. Check access

Complete [Check your setup](/get-started/first-api-call).

**Success:** authentication works, the Echo run completes, and its receipt
confirms `mode: "echo"`, no selected model arm, and
`ranking: "skipped_platform_proof"`.

### 2. Choose model access

Use a model already shown in your catalog, or [use your provider key](/get-started/enterprise).

**Success:** the catalog contains a model you can add to Millwork.

### 3. Run one public task

Complete [Run a model](/get-started/builder).

**Success:** the run reaches `completed`, a result exists, and a receipt exists.

### 4. Check what answered

Compare the model Millwork requested with the model the provider reported.
Check the provider, access method, attempts, and recorded usage.

**Success:** the result and receipt agree about the path that answered.

### 5. Check the boundaries

Confirm that:

* the application request contained no provider credential;
* the task included data, cost, and time limits;
* the receipt did not copy the task or model output.

If a request fails during the evaluation, confirm that it returns safe error
information and a clear next action.

### 6. Decide what comes next

<CardGroup cols={2}>
  <Card title="Continue building" icon="code-2" href="/cookbook">
    Add safe retries, cancellation, key replacement, and receipt exports.
  </Card>

  <Card title="Compare another provider" icon="plug" href="/guides/connect-a-source">
    Connect another account and compare the models it adds to your catalog.
  </Card>

  <Card title="Review the limits" icon="shield-alert" href="/guides/api-sdk-mcp">
    Check whether a requirement is outside the private preview.
  </Card>

  <Card title="Ask a question" icon="mail" href="mailto:support@getmillwork.dev">
    Discuss a requirement without sending credentials or customer data.
  </Card>
</CardGroup>

## Evidence checklist

| Evidence                 | Pass condition                                               |
| ------------------------ | ------------------------------------------------------------ |
| Access                   | Echo completes                                               |
| Model availability       | Catalog contains an eligible model                           |
| Useful output            | Live result contains model output                            |
| Provider identity        | Requested and provider-reported model are visible            |
| Run record               | Receipt contains route, attempts, usage, and checks          |
| Content separation       | Receipt does not contain the task or output                  |
| Recovery, when exercised | A failed request provides safe information and a next action |

<Warning>
  Stop the evaluation if it requires restricted data, a credential in an API
  request, or a production commitment that Millwork does not publish.
</Warning>