Agent Memory
Agent memory is the store that lets an AI agent remember what it has done and learned across steps and sessions: recent conversation, facts it has saved, and outcomes of past tasks, so it does not start every interaction from a blank page.
No jargon, no math
In plain words, what is it?
Agent memory comes in two layers. Working memory is the context window: it holds the current conversation and task, and it evaporates when the session ends. Long-term memory is a database where the agent writes what matters: summaries of completed steps, key facts it extracted (“press 4 uses filter FH-220”), your preferences. Before each new step, the agent retrieves the memories most relevant to the current task and places them into its context, next to your instruction. Continuity is not luck, it is read-back. It breaks in three ways: stale memory (a fact changed but was never updated), context bloat (too much recalled material crowds out the actual task and the model gets worse), and sensitive data written to memory that should never have been stored at all.
Terms that live nearby.
AI Agent
An AI agent is software that pursues a goal on your behalf: it reads the situation, plans the next step, acts through tools (email, calendars, databases, APIs) and checks the result before continuing, instead of answering one question and stopping.
ReadAgent Loop
The agent loop is the cycle an AI agent repeats while pursuing a goal: observe the current state, plan the next action, act through a tool, check the result, then continue, re-plan or stop.
ReadContext Window
The context window is the model's working memory, measured in tokens: everything in a single prompt (instructions, documents, the conversation so far) must fit inside it, and anything beyond the limit is simply invisible to the model.
Read§ 03
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.