Memory Core Guide

What this is

Memory Core refers here to Kiyoraka's Project-AI-MemoryCore, an open-source template for giving an AI companion persistent memory across conversations.

The core idea is simple: use readable markdown files as the AI's memory database. Instead of relying only on one chat thread, the AI reloads a small set of memory files when it is activated.

Why it matters

Most AI chats forget context when the session changes.

MemoryCore shows a practical pattern for continuity:

• identity memory for who the AI is
• relationship memory for user preferences and working style
• current-session memory for temporary context
• optional diary and feature systems for longer-term recall

What to do

Study the project as a memory architecture, not just a prompt.

The main structure is:

master-memory.md
main/
  identity-core.md
  relationship-memory.md
  current-session.md
Feature/
  optional systems
daily-diary/
  optional archive

The important files are:

The activation model is:

[AI_NAME] -> load master memory -> restore core files -> continue with context

The basic commands are:

[AI_NAME]     -> restore AI memory
save          -> preserve progress
update memory -> refresh learning
review growth -> check development

Start with the official repo and read:

• Project-AI-MemoryCore on GitHub
• Setup guide
• Master memory template

Common mistakes

• Treating MemoryCore as one prompt instead of a file-based system.
• Forgetting that current-session.md is temporary session RAM.
• Mixing permanent identity truth with temporary task notes.
• Adding many optional systems before the core files work.
• Assuming the AI remembers automatically without loading the memory files.

Next step

Go to Build Your MemoryCore Agent.

Use the setup guide when you are ready to clone the project, run the wizard, and shape your own agent.