Memory Sets are the containers that bundle an agent's memory profiles so several agents can share the same memory configuration.

Overview

A memory set is a bundle of memory profiles — "Bundled collections of memory profiles" — that you attach to agents so they remember things across conversations. The set itself is just a container with a name, description, and tags; the real behavior (what's remembered, for whom, and for how long) is configured on the memory profiles inside it. The Memory Sets area lists every set in the workspace and is where you create, open, share, and delete them.

Creating a Memory Set

On the Memory Sets list, use New Memory Set (or Add Memory Set from the empty state) to open the create dialog — "Group memory profiles for specific use cases." Fill in:

On create you're taken straight to the set's Profiles tab to add your first profile.

What's on the Memory Set List

Each row shows the Memory Set Name, Referenced Agents (how many agents use the set), Tags, Owner, and Last Updated; the Workspace column appears when you're viewing across projects. Search with Search memory sets. Each row's actions are Edit, Share, and Delete. If you have none yet, the empty state reads Create Memory Set — "Create a new memory set to store and manage your memory profiles."

The Tabs Inside a Memory Set

Open a set to find three tabs:

• Profiles — the memory profiles in this set. Each row shows the profile name, its Memory Type (Semantic, Episodic, or Procedural), Scope, the model used to generate memories, and an enable toggle. Add one with Add Memory Profile.

• Memories — a read-only view of the actual stored memories. Each row shows the Memory Text, Reference Chats (a link back to the conversation it came from), its type, scope, and owner. Search the memory text; click a row to open the source conversation.

• Referenced Agents — the agents linked to this memory set. If none are linked, it reads No Referenced Agents — "No AI Agents are currently linked to this Memory Set."

The set header lets you Share, Publish the set, and Add Memory Profile; it shows Undeployed Changes until you publish a version.

Why Can't I Edit or Delete Stored Memories?

The Memories tab is read-only by design — it shows what the agent has stored, but you don't hand-edit individual memories there. What gets stored, refreshed, and expired is governed by each memory profile's settings (see Memory Profiles), including an optional expiration period that deletes memories after a set duration.

Note: To change what's remembered, adjust the profile, not the individual memory entries. Each stored memory carries provenance — which user's chat it came from, which agent, and which conversation — which is what the Reference Chats link uses.

How Attaching a Set to an Agent Works, and Can Users Opt Out?

Attaching a set simply grants that agent access to the set — its Referenced Agents list is exactly the agents that have been granted access, and at runtime the platform resolves an agent's applicable memories from the sets and profiles it can access. Two enforcement details worth knowing:

• Per-user opt-out — memory can be disabled for a specific user, per profile or per set. An opted-out user's conversations are skipped when memories are written, and no memories are retrieved into their chats.

• Team scope is checked at read time — a team-scoped profile in a shared set only serves memories to users who are members of the profile's team(s), even though the set itself is attached to the agent for everyone.

So sharing a set across agents shares the configuration and store, while scope and opt-outs still control who actually reads and writes each memory.