Privacy Policy for Crabstero

This privacy policy applies to the official Crabstero instance (Discord user ID 746841492617953472) operated by Logan Fick. Crabstero is open source and its source code is available at git.logal.dev. Self-hosted instances are not covered by this policy.

Data Collected

When Crabstero learns from a message sent outside of a thread in a channel it can see, it stores the following data:

• Word pairs derived from message text and embed titles and descriptions.
• Image URLs from embeds.
• The Discord channel ID of the channel the message was sent in.
• The Discord user ID of the user who sent the message.
• Whether a user has opted in to receiving pings via /pingme.
• A minimal opt-out record tied to a user’s Discord user ID if /forgetme is used.

When Crabstero joins a server or gains access to a channel’s message history, it may also process up to 50,000 existing messages from that channel using the same rules.

Crabstero does not learn from messages sent in threads. When Crabstero is mentioned in a thread, it uses the Markov chain of the parent channel.

When Crabstero joins a new server, it also sends a one-time notification to the bot owner containing the server name, Discord server ID, member count, and icon. This information is not persisted beyond the Discord message itself.

How Your Data Is Used

• Word pairs are used to generate Markov chain responses.
• Image URLs are occasionally attached to generated responses.
• Channel IDs scope responses to the channel they were learned in. Data is not shared between channels, even within the same server.
• User IDs are stored alongside learned data so it can be individually identified and deleted upon request.
• Ping preference controls whether Crabstero can ping you in its generated messages.
• Opt-out record prevents Crabstero from learning from your future messages after /forgetme.

Crabstero also maintains aggregate, non-identifying metrics, such as message counts and response times, to help identify operational issues. These metrics are not tied to individual users or channels.

Managing Your Data

You have the following controls over your data:

• /forgetme permanently deletes the data Crabstero learned from your messages across every server and clears any stored /pingme preference associated with your account. To continue honoring your request, Crabstero retains a minimal record of your Discord user ID and opt-out status so it does not learn from your future messages. See Data Deletion on the main Crabstero page for details.
• Deleting a message within two minutes of sending it causes Crabstero to reverse any data it learned from the deleted message.

Data Storage and Retention

Persistent data is stored on self-hosted servers protected by full-disk encryption. Off-site backups are encrypted before upload and stored with a third-party provider, which cannot access their contents.

Crabstero does not use any external analytics or tracking services, and your data is not sold to any third parties.

To support message deletion reversal, Crabstero keeps a temporary in-memory cache of recently ingested message content, embed text, image URLs, channel IDs, and user IDs for up to two minutes. This cache is not written to persistent storage.

Message-derived data is retained indefinitely unless you delete it using the /forgetme command.

Opt-out records created by /forgetme are retained only as long as needed to honor your request not to be ingested.

Changes to This Policy

• 2026-05-04: Initial publication.

Future changes will be reflected on this page.