# Extract a public page

Save one public page with your paid Zilch plan, then check the result before you use it. Use an address you are allowed to read. Do not submit a private address or one that contains secrets.

## Before you start

You need an active paid Zilch plan. Signing in does not grant extraction access. A free Ouros account, another product's subscription, an installed executable or an API key is not proof of a Zilch plan.

Check the subscription shown in your account before you start. These docs are free to read, but extraction is paid. Usage limits apply on top of a plan, not instead of one.

## Extract in the workbench

1. Open the [Zilch workbench](https://dashboard.ouros.md/zilch) and sign in to the account holding your Zilch access.
2. Confirm that paid extraction access is active. If access is missing, review the available subscription options instead of retrying the extraction.
3. Enter a public page URL and select main content or full-page extraction.
4. Start the extraction once and inspect its operation state.
5. Use Markdown only from a completed operation. Check the source URL, warnings and truncation before keeping it.

Hosted extraction does not execute page JavaScript or interact with forms. A successful HTTP response can still contain a failed operation.

## If access is refused

- `401 authentication_required`: sign in with the intended account. This does not activate a subscription.
- `402 subscription_required`: the request lacks the required paid access. Check the subscription on the account; do not switch credentials, install a local tool or submit directly to the API to bypass that restriction.
- A usage limit, or a limit on how many pages you can run at once, means you have reached a cap. It does not tell you whether you have a plan.

Keep the original URL and operation identity when troubleshooting. See the [extraction reference](/docs/zilch/extraction-reference), [account and access](/docs/shared/account) and [recovery guide](/docs/zilch/troubleshooting).

## Follow the exact page request

If the workbench opens in Research, choose **Read a page** for this single-page task. Enter the **Public page URL**, leave **Main content** selected for an article or clear it for surrounding HTML, and choose **Read page** once. Leave **JavaScript** off.

Expand **Page request** and keep the **Operation ID**. If the request is still pending, use **Check existing operation** or **Check selected request**; if only the submission was retained, use **Check existing submission**. Do not create a second request because a wait ended.

In the completed **Page document**, inspect **View extraction details** for `requested_url`, final `url`, `warnings` and `truncated`. Compare a heading and a passage you expected. Only then use **Copy document** or **Download .md**, and reopen the saved file to check that you kept the intended text.

If the document is partial, keep its limitations with your notes and review [warnings and truncation](/docs/zilch/read-truncation-and-extraction-warnings). If the connection drops, use [safe retry recovery](/docs/zilch/retry-without-duplicating-work), not another unrelated extraction.