Agent Memory

Simple Definition

Agent memory is the information an AI agent can keep or retrieve across tasks, sessions, or workflows. It lets an agent carry useful context forward instead of starting from scratch every time.

Types of Agent Memory

• Short-term memory: the current conversation or context window
• Long-term memory: saved facts, preferences, project details, and instructions
• Workflow memory: what the agent needs to continue recurring tasks, like what changed since last week
• Context graph memory: a structured map of how projects, files, tasks, and people relate

Example

A project agent might remember the product goal, previous summaries, coding conventions, and unresolved issues from earlier tasks, so each new request builds on real context.

The Risks

Memory can make agents far more helpful, but it also creates risks. Stored context can become outdated, inaccurate, or sensitive. Worth asking: What does the agent remember? Who can see it? Can it be edited or deleted? Is it still correct?

Related Terms

• Context Window, the model’s short-term working memory
• AI Agent Infrastructure, where memory fits in the stack
• AI Workflow, where memory helps continue recurring work
• AI Agent, the system that uses memory
• Related guide: AI agent infrastructure explained