HollowBot

What it does

When a Hollow user types a URL into the message box, their app requests that page and reads its OpenGraph tags — title, description, and the preview image — so the message can carry a preview card. That is the entire behaviour.

• One request per pasted link. It is not a crawl. HollowBot does not follow links, does not walk your sitemap, and does not come back on a schedule.
• Nothing is indexed or stored on our servers. The title, description and a downscaled thumbnail are embedded in the user's encrypted message and travel to the people they are talking to. We never see them.
• Recipients never contact your site. They render the card from data that arrived inside the message. One paste means one request, no matter how many people read it.
• It requests the page and its preview image. Nothing else.

Where the request comes from

The user's own device, not our infrastructure. Hollow is distributed and there is no server in the middle fetching pages on anyone's behalf, so there is no IP range to block — the address in your log belongs to a person who was reading your page.

robots.txt

HollowBot does not fetch robots.txt, and we would rather say so plainly than imply a compliance we do not implement. The reasoning: this is a single user-initiated unfurl of a link a person deliberately shared, not automated crawling — the same position Slack and Discord take for their own previews. Fetching robots.txt first would also mean two requests to your server instead of one.

If you would prefer HollowBot did not fetch your pages, block the User-Agent, or get in touch and we will help.

Turning it off as a user

Hollow users can disable link previews entirely in Settings → Network → Link Previews. With it off, the app never touches a pasted URL at all.