Weft reads your Messages history only on your Mac and never changes it. It only sends a message when you press Send.
Before a conversation is sent to an AI for the first time, Weft asks, showing which AI, where it goes and how much of the history. To sort, it sends the text of the conversations you added (and who sent each message) to the AI you chose, and nothing else. With Ollama or LM Studio, message processing stays on your Mac.
Weft has no account, no server, no analytics and no tracking. At launch it contacts GitHub to check for updates and to read its model recommendations; GitHub sees your internet address, and no messages or personal data are sent. Notifications can show message text on the lock screen, so there is a setting to show only “New message” instead.
Contacts access is optional. Weft shows names next to numbers, and the names are also part of the text sent to the AI you chose, so it knows who said what. Without Contacts access, phone numbers and email addresses are sent instead. With Ollama or LM Studio, message processing stays on your Mac.