Import

Already have help articles, documents, or support emails? Import (import from your website or files; some people call it upload or smart import) is the fast way to get them into your Knowledge Base. You point it at a source, and the AI reads your content, cleans up the formatting, merges duplicates, and files it away - ready for your agent to use.

How It Works

Every import follows the same three steps:

1. You pick a source - upload files, or connect Google Drive, Notion, a website, or your email.
2. The AI processes it - reading, cleaning, and organizing each item in the background.
3. It lands in your Knowledge Base - ready for your agent to search.

Import from Files

Uploading files is the simplest way to start. It accepts PDF, Word, Excel, HTML, CSV, Markdown, text, ZIP archives, and code files.

Import from Other Sources

The other sources work the same way - select the source on the Import page, connect it, and click Import knowledge:

• Google Drive - connect your account (read-only - DeskClone never changes your files) and pick a folder. Everything in it, including subfolders, is imported. Google Docs and Sheets are converted automatically.
• Notion - connect with an integration token, then choose pages or your whole workspace.
• Website - type your website's address (yourstore.com is enough). DeskClone reads that page and follows its menus and footer links to pull in related pages too (up to 50 pages, 3 levels deep) - great for importing an existing help center. If your site sends visitors from yourstore.com to www.yourstore.com (or the other way), it follows, and keeps using that address for later syncs. It never reads pages that aren't for customers: account and login pages, the cart, checkout and search. If nothing could be read, the import says why and what to try.
• Email history - choose a connected Gmail account and a date range. The AI distills each thread into a clean article (summary, key details, people, and resolution) and drops the reply-chain noise. This is different from email monitoring, which creates tickets from new mail.

Requesting a Source We Don't Support Yet

Some sources in the left-hand list - Intercom today - aren't built yet. Selecting one shows a card that says so plainly, along with a Request [name] integration button. Clicking it tells us you want it; we use these requests to decide what to build next. It does not start an import - there's nothing to import until we build the connection.

What It Counts Toward

Import is not charged per import. DeskClone bills on monthly conversations, not on imports. Each plan instead includes a monthly file allowance - 50 files on Free, 300 on Starter, 1,000 on Pro, and unlimited on Business - and every file you import counts toward it. A website import counts only the pages it actually reads, and when your allowance is nearly used it reads up to what is left. So bring in as much content as your plan's allowance allows; importing more never adds to your bill.

Watching Progress

After you start an import, the Jobs tab shows each job move through pending → extracting → synthesizing → done, with a per-file view of progress. Large imports take a few minutes; you can leave the page and come back - it keeps running in the background.

When a job finishes, the content is added to your Knowledge Base automatically. The Where files came from tab shows which original files built each article, so you always know where an answer came from.

What the import decided

A finished job also shows What this import decided - the calls it had to make about which article a document belongs to. Nothing has gone wrong when you see these; they're choices only you can confirm:

• Worth a look. Two articles cover similar ground, so both were kept and both are live. If one is out of date, delete it - your agent reads both.
• Updated, not added twice. The same document arrived under a new name, so it updated the article you already had instead of creating a second copy.
• Your own file was kept. An import matched a page you wrote yourself. Yours is untouched; the imported version was saved alongside it.
• Your edit was kept. You changed a page in your knowledge after the last sync, so a re-sync did not overwrite it with the newer version from your website. Yours stays, and later re-syncs keep it too. This is a choice the sync made for you, not an error.
• Kept in this store's folder. An earlier read of a connected store kept one of its pages outside the store's own folder. It now lives in the folder, so only that store's agent reads it.
• Archived, not deleted. The page is no longer on your site, so it stopped answering. Nothing was removed - restore it from your Knowledge Base if it should still be there.

Re-importing the same source always updates the articles it created last time, so your knowledge stays current without collecting duplicates.

Keeping Website Knowledge Up to Date (Sync)

Once you've imported a website, it appears under Your knowledge sources at the top of the Jobs tab, with how many pages it covers and when it was last synced. Click Sync now to re-read the site and update anything that changed - your agent's answers stay current with your latest pages.

Sync is built to be safe and cheap:

• Only changed pages update. Pages that haven't changed are skipped entirely - they don't count toward your file allowance, so re-syncing an unchanged site costs you nothing.
• Your organization is preserved. Files you've renamed, moved, or made Team only keep those choices. A file you've edited yourself is never overwritten by a sync - your version wins, and the card tells you so.
• Removals are careful. A page that's genuinely gone from your site is archived (recoverable), but never when the site was only partially reachable - a bad day on your website can't wreck your Knowledge Base.
• Big changes wait for you. If a sync finds sweeping changes (say, after a site redesign), it's held with a Waiting for your review badge instead of applying automatically. Open the job, look it over, and choose Apply changes or Discard update.

Each source can be synced once per hour. If your knowledge hasn't been synced in a while, the dashboard's Knowledge card reminds you with a one-click Sync now.

Automatic Updates

You don't have to remember to sync. On the source card, click Turn on automatic updates - we'll check your website monthly and update it when things change. If we find big changes, we'll ask you first. Change this anytime.

What to expect:

• The first check emails you a receipt - even when nothing changed ("everything is already up to date"), so you know it's working.
• Big changes wait for your review. The source card shows Automatic updates waiting on your review, you get one email with a link to the review screen, and your agent keeps answering from your previous content until you decide. Automatic checks pause while a review is waiting and resume when you apply or discard it.
• Silence means all is well. Successful checks don't email you - your weekly summary carries one line ("Your website knowledge: checked this month, 2 pages updated") instead.
• If your site is unreachable for two weeks, automatic updates pause and you get one email saying why and what to do. Your knowledge is unchanged; turn them back on from the source card when the site is healthy.
• Only changed pages count toward your monthly Import allowance - a check of an unchanged site costs nothing.

Automatic updates come with Starter and above.

Handling Imports From the Chat

Jarvis can look after imports from the chat. Ask it to show your imports, apply one that is waiting for your review, turn automatic updates on or off for a source, or sync a source now.

• It applies only what the preview showed. The files it writes are exactly the ones in the import's preview, and you can name which to apply or take them all. Applying closes the import, so a file you leave out is not applied later; name every file you want.
• Removals stay your choice. If pages are gone from your site, Jarvis never removes their files. It tells you how many were left, and those pages are proposed for removal again at the source's next update.
• Your edits are kept. A page you edited yourself keeps your version, and the result says so.
• Removing and reverting are cards. Jarvis can remove an import from your history, revert a whole import (it deletes the files that import created), or revert one change it made (a created file is deleted; an updated one gets its earlier text back). Each puts a card on your screen naming the import and the files, and nothing happens until you tap. None of them has Undo. It reaches only imports whose files are in clients you can open.
• Undo is careful. Undo restores the files the import updated and deletes the ones it created, except any you changed since. Those stay as you left them, and it names them.
• Automatic updates need Starter or above, as above. Jarvis can check daily, weekly or monthly; the dashboard's button sets monthly.
• Sync now is once an hour for each source. It starts the update and returns at once; the update takes a few minutes, and Jarvis will not say your knowledge is updated until it is. If a sync is already running, or the source was synced within the last hour, it says so.

What's Next

With your Knowledge Base filling up, you can teach your agent to take actions too - head to Tools. Or, if you just want to go live, jump to Add Your Agent to Your Website.