Facts
People and preferences
Save details about yourself and people you know, then ask for them when they matter.
“Remember Sarah likes hiking.”
Memory and the brain
Save a detail, remember a place, or return to a conversation. Avenkin's memory lives on your device, so useful context can come with you when you change AI models.
What it remembers
Memory brings together several kinds of saved information. What you can recall depends on what was saved or captured and which features you enabled.
Facts
Save details about yourself and people you know, then ask for them when they matter.
“Remember Sarah likes hiking.”
Places
Remember where you parked or left an object. Location can be included when you grant permission.
“Remember my car is in lot B, level 3.”
Conversations
Search saved conversations for earlier decisions and details. Meeting summaries use a captured transcript, such as ambient captions enabled before the meeting.
“What did we decide about the project last week?”
Knowledge
Search available notes, documents and remembered facts together, with source references when available.
“What do I have saved about this project?”
Follow up
Record what someone asked for or what you owe them, list open follow-ups, and mark them done.
“Remember I owe Bob a copy of the deck.”
Rewind
With memory rewind enabled, use the opt-in rolling audio buffer to revisit recent speech. It cannot recover audio from before you enabled it.
“What did they just say?”
The brain, explained
The brain is Avenkin's on-device system for searching saved context and connecting related information. For example, a saved relationship can link Sarah to the company she works for, while a follow-up records the document you promised to send her.
When you ask a question, Avenkin can search across those sources and provide relevant findings to your selected AI model. Conversation recall can return source references, so you can check what an answer is based on.
The AI model generates the reply; the memory supplies context. Changing your selected model does not move your saved memory into that provider's account. Retrieved information may still be included in a request to that model.
1
“Sarah works at Acme. I owe her the project notes.”
2
A saved person, a company relationship and an open follow-up provide context for later questions.
3
“What do I need to send Sarah?” retrieves relevant saved context for the assistant to answer.
Storage and processing
Memory storage and AI processing are separate choices. A cloud model may receive relevant recalled facts or transcript excerpts with your question. On-device AI options include Apple Intelligence on supported devices and compatible downloaded models running with MLX or llama.cpp. Speech recognition and spoken voice are configured separately; connected tools and exports have their own destinations.
Apple Intelligence needs no API key when its system model is available. It supports text conversations and selected enabled app tools, including Field Assist job actions, calendar, reminders and calculations. The tools offered depend on enabled features and the active session. Camera understanding still needs a vision-capable model. For an entirely local conversation, also choose local speech and voice and disable internet-dependent tools.
Read the privacy policy or follow the offline setup guide.
Your control
Open Memory to search saved facts and inspect their sources. Correct supported entries, or ask what Avenkin has saved about a person or topic. Inferred information can be wrong.
“What do you remember about me?”
Ask Avenkin to forget a saved fact, or use Memory on your device. The forget flow asks for confirmation and explains where a copy may remain.
“Forget that Sarah likes hiking.”
Conversations and recordings are kept separately. Forgetting a fact does not by itself erase the source conversation, an exported file or a copy already sent to a connected service.
Save one useful detail, ask for it later, then review it in Memory. The memory and recall guide has more examples and setup notes.