Right context per turn
Toggle Vault on, Pinned only, or Off so a private question stays lean while a project question gets full recall.
Chat
Chat memory modes
Dial vault context up or down per turn without losing the record.
Overview
What Chat memory modes does
Memory modes decide how much of your vault reaches the model on a given turn. Vault on pulls the most relevant entries plus your pinned memories into the system prompt; Pinned only narrows context to the memories you have marked as always-active; Off attaches no vault memory at all. You switch modes from the chat header, so sensitive or focused turns get exactly the context they need.
The modes are a budgeted injection, not a firehose. Explicitly attached entries and files always win, pinned notes come next, and relevant search results fill the remainder within a token budget so the prompt never overruns the context window. Whatever mode you choose, the conversation still saves back into the vault, so turning memory off for one turn never means losing the record.
Memory modesContext
Why it matters
The payoff for your AI memory
Pinned memory stays reliable
Marked memories are always available in Vault and Pinned modes, giving the model a stable base of the facts you care about most.
Budgeted, never bloated
Context is trimmed to a token budget with explicit attachments first, so injected memory stays relevant instead of overflowing the window.
How it works
From first launch to reusable memory
- 1
Open a conversation and find the memory-mode control in the header.
- 2
Choose Vault on, Pinned only, or Off for the next turn.
- 3
Send your message and let the chosen amount of context attach.
- 4
Adjust the mode any time as the conversation shifts topics.
FAQ
Common questions
What do the chat memory modes do?
Vault on adds relevant and pinned memories, Pinned only restricts context to your pinned entries, and Off attaches no vault memory — letting you control per-turn context.
If I turn memory off, is the chat still saved?
Yes. Memory mode only controls what context is injected into the prompt; the conversation is still persisted to the vault regardless of the mode you pick.
Related
More Chat features
Unified Chat
Chat with the vault inside 1AIVault and route turns to local coding agents without leaving the app.
Learn moreAgent routing and model picker
Detected local agents expose available models, setup errors, and route choices before a message is sent.
Learn moreAgent, vault, file, and skill mentions
The composer can mention agents, vault entries, topics, local files, folders, and skills as reusable context chips.
Learn moreLocal memory, shared everywhere
Give every AI tool the same memory.
Start free, import real conversations, and reuse your memory across every AI agent you already use.