Context is what the AI agent can read right now in its active session — the instructions, the conversation, the documents currently loaded. Memory is information stored outside the session that can be retrieved and added to the context when needed. They are different systems that serve different purposes.
Context: The Working Desk
Think of context as everything currently on your desk while you are working. Your notes, the document you are editing, the email you just opened — all of it is right there, immediately accessible. This is the context window. The agent can see and use everything on that desk. But the desk has a fixed surface area. When it fills up, something has to move to make room for the next thing.
Context is fast and immediate. The agent does not have to look anything up — it is all right there. But it only lasts for the current session. When the session ends, the desk gets cleared. The next session starts with a fresh desk, and whatever was on the old one is gone unless you specifically reload it.
Memory: The Filing Cabinet
Memory is everything in the filing cabinet beside the desk — organized information that persists between sessions and can be retrieved on demand. An AI agent with a properly designed memory system can pull relevant files from the cabinet and place them on the desk when it needs them. The student policy document, your past interactions with a specific learner, your course outline — these can all live in memory and be fetched as needed rather than sitting on the desk taking up space all the time.
Different AI tools implement memory differently. Some store it as structured files the agent reads at the start of each session. Others use vector databases that retrieve the most semantically relevant chunks for each query. Claude Code and similar agentic frameworks use files like CLAUDE.md as persistent memory — documents the agent re-reads at the start of every session to restore essential context.
What This Means for Educators
For coaches and consultants building campus AI agents, understanding the context-memory distinction helps you design better systems. Essential instructions that the agent needs every session — your program rules, your audience profile, your communication style — belong in memory and get reloaded as context at the start of each session. Student-specific information, historical interactions, and reference documents belong in a knowledge base retrieved on demand. This architecture keeps the active context clean and focused while giving the agent access to a much larger pool of information than any context window could hold.
The Simple Rule
Context is what the agent can see right now. Memory is what the agent can retrieve when it needs to. Design your agent so its context always contains what it needs for the current task, and its memory contains everything else it might ever need to look up. Keep the desk clear. Keep the filing cabinet organized.