Data storage

This page explains, from a security perspective, where Keragi data lives, who can decrypt it, and what happens when data is deleted. Database choices and deployment configuration (SQLite by default, plans for other databases) are covered in Runtime and database.

1. Server relay, client storage

Keragi uses a server relay, client storage model by default:

2. What the server stores

The server database mainly holds the non-plaintext data required to operate the service:

The server may process metadata related to accounts, connections and routing, so relay mode is not anonymity.

3. What the client stores

The client stores the material that can decrypt messages:

LocationMain contentsCan it decrypt messages?
ServerAccounts, public keys, conversation indexes, ciphertext messages, connector and resource stateNo
ClientPrivate keys, conversation keys, plaintext messages and local history, client settingsYes

4. Risk of deleting client data

Important: Messages, keys and similar data ultimately live on the client. Clearing browser site data, uninstalling the client, resetting a device or manually deleting client data may also delete the messages and keys stored on that device. Without another signed-in device or a usable backup, those messages may be unrecoverable or impossible to decrypt. The server only holds ciphertext and cannot recover them for you.

Recommendations:

5. Backup and recovery

6. Data path for AI and connectors

AI takeover is client-side logic that decides whether to reply based on conversation settings. When you enable a third-party AI service or connector, relevant conversation content is sent to that service under your configuration; its processing and retention are controlled by that provider and are outside the end-to-end encryption boundary.