Add Claude a memory that it REALLY uses

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Working with Claude has become very frustrating
You need to implement a way to mantain some certain memory between chat changes, and when context limit is reached.
It is VERY frustrating to work with a tool that memory is cleared from time to time, asking nonsenses, proposing development that is working for weeks, making fixes on things that are already in production, proposing milestones that already off the project months ago... every change of context is a birth... uéee... uéee...
Claude.md is made for WHAT??? All the information thtat we hardcode there, is just for looks.
It is apologize for this, apologize for that, im sorry i did this, i shouldnt had done that... all the time... Claude become a apologetic agent that really DONT GIVE A FUCK about you clearly hardcoded as guardrails. It seems to me that Antrophic is making money with all that comes anf goes of rewriten information all the time... tokens are payed no?
MAKE SOMETHING to resolve that... cannot be so difficult... we need a permanent memory that is REALLY present all the time... this is totally unaceptable... i pay to have a AI that helps me not to have a employe that i have to teach how to work every 10 minutes... we dont need a huge permanent context, but something that points to a minimum quantity of information that leeds Claude to have a minimum understanding of the project, everytime the context change... creating a huge context before restarting a chat, help with the task, but does nothing to the vision

Proposed Solution

Create a document that Claude have on memory all the time with a minimum context, to guide the development memory.
Simple like

• guardrails: look document A
• product PRD: look for document B
• readme: look for document C
• development Log: look for document D
• project memory: see document E
• on every start: read document F

Alternative Solutions

Make Claude REALLY read Claude.md on every context change.

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

Everytime the context is over Claude thinks that is a new project. Forgets development and try to redo things that already done.

Additional Context

Excuse this
Excuse that
I shouldnt have done this
Forgive me, i shouldnt have done that
All the time