Memory and Knowledge

Aori’s memory system is designed to give the agent persistent, personal context without requiring you to maintain it manually.

Palace Memory Architecture

Aori uses a structured memory system called the Palace. Instead of a flat log, the Palace organizes what Aori knows into named memory sections — similar to rooms in a building. Each section holds a category of knowledge (your preferences, ongoing projects, key facts, and so on).

Memory is extracted automatically from your conversations and managed internally by Aori. You don’t need to maintain it — it’s updated as you use the app.

How Memory Works

1. Extraction: After conversations, Aori identifies facts worth remembering — names, preferences, commitments, recurring topics.
2. Organization: Facts are filed into Palace sections (e.g. “About Me,” “Projects,” “Preferences”).
3. Recall: When you start a new conversation, Aori retrieves relevant sections to provide context automatically.

You do not need to tell Aori to “remember this” — it handles extraction automatically. For things you want explicitly remembered, you can say “remember that…” and Aori will prioritize it.

Interacting with Memory

All memory interaction happens in chat. Ask Aori what it remembers, or tell it to forget something. You can wipe all memories at any time by asking Aori to do so.

Knowledge Sources

In addition to conversation memory, Aori can draw on structured knowledge you connect as Skills:

• Local Folder — Index files from a folder on your device
• Obsidian Vault — Use your Obsidian Markdown notes as a knowledge base
• Notion — Search your Notion workspace in real time

See Skills Overview for setup details.

Data and Privacy

All memory data is stored locally on your device. There is no Aori cloud for your memory data. If you use a cloud AI provider, relevant memory sections may be sent as context for a given conversation — only what is needed for that request.