Skip to main content
Collect readable pages from a public website and save them as knowledge for your agent. A scan finds pages. Saving their content to memory makes it available for answers. This does not install a chat widget, publish the agent or give it unrestricted live web browsing during conversations. Use an agent you can edit and a website you own or are allowed to use. Choose a public starting page you can read without a customer login. For a first scan, a clear FAQ or service page is easier to check than a large site with unrelated material.

​
Open the website setup

  1. Open Home.
  2. Select the intended agent.
  3. Open Build → Settings.
  4. Find Knowledge → Website.
  5. Select Add your website.
Home opens your current space; you choose the agent there. If the setup dialog is already open, continue with Choose the address and scope. If the intended website is already attached, select Review pages on its source card and continue with Review the page results. To keep that source while adding another starting address, use Add a website through Library.

​
If the form is unavailable

If an upgrade explanation appears, review the offer before choosing Upgrade to Pro. Not now closes it. You do not need to buy a plan just to finish a practice example. You can use Upload files to your agent or Paste text as knowledge for approved document material instead. If Add your website is missing, check the intended agent’s Build → Settings → Knowledge area. An existing website has a source card instead of a second new-source button. Use Review pages, then continue with Review the page results. If the Website section is absent entirely, use Get help from a person to ask about availability for your account. Do not assume the site was added.

​
Choose the address and scope

Paste the complete address of your own public starting page into the website-address field. It may be labelled Your website address. Use the page you actually want collected, rather than an example company’s address. Select Crawl settings to reveal optional controls. A crawl is a scan that follows website links. These choices affect the collected material, not the agent’s tone or intelligence. The offered fields, labels and bounds can vary. Follow your displayed form. These are the purposes of the controls you may see: For Pages to skip, use paths from your own site. For example, enter /blog and press Enter or Space to make it a visible entry. This excludes that section and its pages. You can also paste a full link from the same site. Use an entry’s small remove button to undo it. This changes collection, not the original website.

​
Set up a small first scan

The following are practice choices, not product defaults or a promised page count. Use them only where the fields and values are offered:
  1. Enter your public FAQ or service-page address.
  2. Open Crawl settings.
  3. Set link depth to 1, if allowed.
  4. Set maximum pages to 10, if allowed.
  5. Leave Pages to skip empty unless you know a section should be excluded.
  6. Choose Never for re-scanning, if offered.
Never lets you refresh manually. A recurring option such as Weekly, when available, requests later scans. Neither a schedule nor a scan guarantees a website edit is available to answers the moment you publish it.

​
Start the scan

  1. Check the address and optional scope settings.
  2. Select Scan website.
  3. Wait for page results.
Correct any marked field before selecting Scan website again. The source form does not use Design’s Save Changes.

​
Review the page results

Read the page names and addresses before saving them as knowledge. The header distinguishes pages found from pages saved to memory. A page marked Scanned was found but is not yet saved knowledge. Expand a page to inspect its content when needed. Use Remove for a page you do not want included. If you are back on the website source card, select Review pages to open the list, then continue with Save the pages to memory.

​
Save the pages to memory

Save to Memory handles eligible pending or failed pages; it is not a checkbox-selection step. Remove unwanted pages first.
  1. Select Save to Memory.
  2. Wait for the pages to finish saving.
  3. Check that the intended pages show Saved, rather than only Scanned.
For one pending page, use its row’s Save action, then check that it becomes Saved. The global button may be unavailable while other work runs or when no eligible content remains to save. If the intended pages already show Saved, continue with Check Active and try a question.

​
Check Active and try a question

The website source stays attached to this agent after setup. On its source card, check the switch beside Active or Paused. Turn it on if you want the agent to use the saved pages; it should show Active. This switch controls use of the knowledge. It is separate from the re-scan schedule. A success message may say Your agent now knows your website — ask it something. Select Ask in the preview to send its suggested question. To ask your own, use the chat beside the Build settings or select Chat Preview within Build on a narrow screen. Enter a question about a fact you just checked on a Saved page, then select Send message or press Enter. Compare the reply with that actual page. Saved, active knowledge gives the agent material to consult; it does not guarantee every answer is correct.

​
Recover an empty scan or failed page

  • No useful pages appear: reopen the starting page in your browser and check its visible text. Some sites do not provide readable content to the scanner. Use approved uploaded or pasted material if needed.
  • Useful content is several links away: review the scope using Change website scan settings for an existing Library website.
  • A page shows Failed: read its message, correct the cause and use that page’s Retry control.
  • A page only shows Scanned: complete Save to Memory or its row’s Save action.
  • The original site changed: select Refresh on the source card, then review the new results and saved state before expecting the edit in answers.
Refresh uses the saved address. It does not choose a new starting page. To retain this source while adding another address, use the Library path below. Avoid repeatedly adding the same website because one page has not finished.

​
Add a website through Library

Use Library to organize website material centrally. You need permission to add a website in the chosen Library project.
  1. Open Library.
  2. Select Add new.
  3. Select Add your website.
  4. Enter your public starting-page address.
  5. Select Add website.
  6. Read the result before continuing.
This dialog uses Add website, while the direct-agent form uses Scan website.

​
Check the Library scan and pages

When the scan starts, the dialog closes and Alysium explains that pages will appear as they arrive. A saved website entry alone does not mean its pages have been collected. If the website was saved but its scan could not start, the dialog stays open with a message explaining why. A plan refusal may offer Upgrade. Read that offer before deciding; you can keep the saved website without buying a plan. Continue with Recover the existing Library website. For a scan that started:
  1. Open the existing website entry in Library.
  2. Read the scan status beside its heading.
  3. Check the Pages list as results arrive.
  4. Open a relevant page to check the collected information.
While the scan is running, wait for its page results. If it fails or finishes without the needed material, read its status and any displayed message, then continue with Recover the existing Library website. Once the needed pages are available, continue with Attach the Library material.

​
Recover the existing Library website

Read the scan refusal or error and correct its cause. If Upgrade is shown for a plan restriction, review the offer before choosing a purchase. Uploaded files or pasted text are alternatives for approved material. For missing pages, check that the original public page contains readable text. Use Change website scan settings if the collection scope needs to change. After correcting the cause:
  1. Close the add-website dialog if it is still open.
  2. Open the saved website entry in Library.
  3. Select Scan again beside its heading.
  4. Read the result.
  5. Return to Check the Library scan and pages.
Recover the existing entry rather than adding the website again. If the restriction remains unclear, use Get help from a person with the address and exact message.

​
Attach the Library material

After checking the collected pages, follow Use Library files in an agent to attach the material through Add from Library. Once attached, use the intended agent’s Build preview to ask about a fact you checked. On a narrow screen, select Chat Preview within Build. Enter the question, then select Send message or press Enter.

​
If an answer misses website information

Check the relevant page is listed and Saved, and the source is Active. Review Knowledge → Behavior, then ask a direct question about the page’s actual contents. Why your agent can’t find something supplies further checks. Control how your agent uses its knowledge explains grounding and missing-answer choices. To put a chat window on the site, use Put your agent on your website. That separate Deploy task requires access to the website’s editing tools.