Memory

Memory is what the assistant can remember about you and your work. It is organized as memory files you write or upload, automatic per-chat summaries, and an optional knowledge graph, all of which you view and manage. This page explains what the memory manager shows, the controls you set, how a file's state changes the way it is used, and how the privacy toggles work. Memory is part of the AI Console and is available on the Free, Pro Lite, and Pro plans, with storage and pinned size capped per tier.

Before you begin

Sign in to your account. Your memory is private to you.
Open the memory manager from the sidebar, or, on desktop, from the panel pinned beside the chat.
To understand how saved documents are searched and recalled, see Your knowledge and retrieval.

Quickstart

Open the memory manager
Go to the AI Console and open Memory from the sidebar.
Add a text memory
Select New text memory, then write the preferences, rules, or context the assistant should always know.
Decide how it is used
Leave a file active and unpinned to have it pulled in only when relevant, or pin it to include it in every chat. Keep pinned files small.
Check the meter
Watch the storage and pinned-tokens meter as you add content, and trim or unpin if you near a limit.

What it shows

The memory manager has three tabs, plus a scope switch and a usage meter. The areas are described in the following sections.

Memory files

A file browser on the left and an editor on the right. Two kinds of entries appear:

Text memories you write directly, such as preferences, rules, watchlists, or standing context.
Uploaded files whose text is indexed so the assistant can find them by topic.

A system-maintained AI Journal entry sits pinned at the top of the list. The assistant appends a one-line fact to it when you say "remember X" in a chat, and the entry links back to the chat that produced the last line.

Each file has two states you control:

Active or Disabled: a disabled file is ignored.
Pinned or unpinned: a pinned file is included in every chat, so keep pinned files small. An unpinned active file is pulled in only when it is relevant to your question.

Chat memory

Automatic short summaries of your past chats, so older threads stay useful from new ones. You can pin a summary to always include it, pause a chat to freeze its summary, or make a chat incognito to skip summarization entirely.

Knowledge graph

When this view is enabled for your account, it shows the entities and facts the assistant has noted about your work, with a link back to the chat each came from.

Scope and meter

A scope switch on the files tab lets you view your main pool, a specific subagent's private memories, or everything at once. A storage and pinned-tokens meter appears as you approach your limits.

Inputs

The controls in the manager are described in the following list.

New text memory: creates a blank file. Write up to about 20,000 characters of whatever the assistant should always remember.
Upload memory file: adds a document whose text is indexed for retrieval.
File name and Category: organize and label a file.
Description and When to use (for uploaded files, up to 1,000 characters each): help the assistant decide when to pull the file in. A Regenerate with AI action can draft these for you, with optional steering text, for you to review and save.
Active toggle, Pin toggle, Move to (move a file between your main pool and a subagent's private collection), and Delete.
Search: filter files by name or content.

The four privacy toggles live in Chat settings and all default on. The wipe action lives in the same privacy area.

Auto-capture memories: the "remember X" tool that appends to the AI Journal.
Auto-accumulate chat memory: the rolling chat summaries.
Allow portfolio access: lets the assistant read your saved portfolios.
Allow brokerage access: lets the assistant read the holdings from a brokerage account you have securely connected.
Wipe: a confirmed, destructive action that clears your stored memories, chat summaries, and knowledge-graph entries.

Turning a toggle off stops that writing or reading. It does not delete existing data.

Outputs and how to read them

File state: a pinned, active file is included in every chat. An active, unpinned file is retrievable when relevant. A disabled file is never used. The editor footer states which state a file is in.
Chat summaries: keep your earlier conversations searchable from new chats. A summary refreshes after an assistant turn once the chat has at least five new messages or about five idle minutes.
Storage and pinned-tokens meter: warns you as you near your limits, so you can trim or unpin.
Recall reference: in a reply, a recall line can link back to the exact memory file the answer drew from, so you can open and check the source.

How it is calculated

Not applicable. Memory is content you create and the assistant maintains. Retrieval of an unpinned file is done by topic matching against the file text, its description, and the when-to-use note you provide. No financial calculation is involved.

Data and timing

Memory holds your own content: notes, uploaded files, and summaries of your own chats. Nothing here is market data. A chat summary updates shortly after a qualifying turn, once the chat has at least five new messages or about five idle minutes. The assistant reads your memory alongside live data only when it answers a question.

Use cases

Save your investing preferences and risk rules so every chat respects them.
Upload a research document and have the assistant reference it when relevant.
Keep a default portfolio or watchlist the assistant assumes when you ask about positions.
Make a sensitive thread incognito so no summary is stored.

Limitations and disclosures

Memory storage and the number of always-pinned tokens are capped per tier, with the limits surfaced in the meter. The memory surface itself presents no valuation, forecast, screen, or portfolio result, so it carries no advice of its own. The standard not-investment-advice line still applies to any chat answer that uses your memory.

GNG Research provides equity research and educational tools, not investment advice. Nothing on the platform is a recommendation to buy or sell any security. Do your own research and consider your circumstances before making any investment decision.