Also known as: persistent agent context, agent recall
Agent memory is the persistent context an AI agent retains across conversations and runs. Unlike a stateless chat, an agent with memory remembers past decisions, client preferences, project context, and operational patterns. It picks up where it left off instead of re-briefing every session.
Where you'll see agent memory in day-to-day work inside Helm.
In Helm, agent memory is workspace-scoped and permission-respecting. An Account Manager agent remembers that Acme prefers Friday check-ins, that their primary contact is on PST, and that the last three conversations were about the upcoming rebrand, because it wrote those to memory during prior runs. Memory is stored per agent and read only by that agent. There's no cross-client leakage because memory is attached to the same records the agent already has access to.
Concepts that show up in the same workflows and reports.
Common questions and honest answers.
No. Conversation history is the log of a single session. Memory is structured knowledge the agent has chosen to persist, like 'Acme's Q2 budget is fixed-fee' or 'Jordan at Globex prefers async updates, no meetings'. Memory survives across runs.
Yes. Agent memory is visible to workspace admins and the agent owner, and entries can be deleted or edited. This matters for GDPR compliance and for tuning when an agent learns something wrong.
No. Memory respects the same row-level security as the rest of the platform. An agent scoped to Client A cannot read memory written about Client B.
Helm is the AI work platform where these concepts stop being theory and start being your Monday morning.