What is agent memory? Give AI useful continuity
Memory gives an agent continuity without forcing a person to repeat the same context in every prompt. AI Agent lets each agent use recent messages, semantic recall, structured working memory, and observations where they help.
memory
Weekly Growth Audit
remembered
7
open loops
2
Run history
Last pass found activation leaking after workspace invite.
Pinned preference
Keep reports concise and bias toward measured deltas.
Operating context
Q2 sprint, onboarding focus, Monday exec update cadence.
Next run starts with
3 remembered decisions, 2 open questions, and the last recommended onboarding experiment.
The concept
A plain-language definition before the product implementation.
Memory is selected context from prior work
An agent does not remember like a person. It stores prior messages or structured notes, retrieves the pieces that fit the current thread, and places them into the model's context. Recent history handles continuity, semantic recall finds older relevant exchanges, working memory keeps compact facts, and observational memory condenses longer-running activity.
How AI Agent uses it
Configure memory per agent
AI Agent exposes memory settings and contents in the agent inspector. Turn recent history, semantic recall, working memory, or observational memory on where the job benefits, inspect what has been stored, and keep memory scoped to the agent's thread or resource contract.
Recent history
Carry the latest messages into the next turn for conversational continuity.
Relevant recall
Retrieve older messages by meaning when the current work relates to them.
Compact state
Keep preferences, goals, decisions, and observations without replaying every raw message.
From idea to a working system.
- 01
Configure
Choose the memory modes that fit the agent's job.
- 02
Accumulate
Store messages, structured facts, or observations as work happens.
- 03
Recall
Supply recent or relevant context to a later run and inspect the stored contents.
Carry forward useful context, not every message
Best for
Recent conversation continuity, relevant prior exchanges, preferences, goals, decisions, and observations accumulated through work.
Choose another pattern when
Authoritative policies or product documentation. Use managed knowledge when the source itself must remain the ground truth.
Included in AI Agent
The product capabilities behind the idea.
- Recent message history
- Semantic recall
- Structured working memory
- Observational memory
Frequently asked questions
Keep exploring
Related product guides
Knowledge
Knowledge gives an agent a trusted source set beyond its model training. AI Agent turns approved documents and websites into retrievable context, then records what the run actually used.
Agents
An agent is the reasoning layer: it understands a goal, decides which approved tools to use, and works toward an outcome. AI Agent gives that reasoning a clear operating contract instead of relying on one giant prompt.
Chat
Agent chat is the conversational control surface for real work. Ask a question, attach context, choose an agent and skills, then review the evidence, tool activity, and proposed changes in the same thread.
Build the smallest useful version first.
Start with the agent and one real task. Add workflow control, context, and approvals when the work shows you where they matter.