You re-explain your market, your price points and how you write at the start of every single chat, and you assumed that is simply how the thing works. It is a setting, it is two switches on one screen, and there is a third screen where you can read and edit everything it has stored about you.
A short clip aimed at agents whose assistant "feels like it forgets you every single time you open up a new chat".
The fix is three steps, all inside settings:
"Go to Claude settings and click capabilities."
"You're going to see two toggles. Turn on memory, then turn on search and reference, and make sure you click both of them."
"Then click view and edit memory, and you can see everything that Claude has stored about you."
With both switched on, the claim is that it holds "your tone, your market, your climate, and your preferences" across chats instead of starting from nothing every time.
And the memory is editable. You can add to it, change it, or tell the assistant directly what you want stored and it updates immediately.
“Go to Claude settings and click capabilities. You're going to see two toggles. Turn on memory, then turn on search and reference”
The reason your AI output reads generic is usually not the prompt. It is that every chat begins with an assistant that has never heard of your market, your price bands, your brokerage's rules or the way you write. Most agents solve this by pasting the same paragraph in every time, which works, and which nobody keeps doing past week two.
The editable memory screen is the part worth an actual half hour. Open it once and write in the things you would otherwise retype: your market and its neighborhoods, your usual price range, the compliance language your brokerage requires, and how long you want an email to be. Then look again in a month and delete whatever it inferred about you that is wrong, because it will have inferred some things.