Overview
The Memory tab in Agent Builder → Settings → Memory shows all persistent memories stored by your agents across conversations. Agents use memory to remember user preferences, prior decisions, and accumulated context — so they can pick up where they left off.
How Agent Memory Works
When an agent decides something is worth remembering — a user preference, a decision made, or a key piece of context — it stores it as a key-value pair in the memory table. On subsequent conversations, the agent retrieves relevant memories and uses them to inform its responses.
Memory is stored per entity (typically a user ID or session identifier) and per memory type. The table shows all memory entries across all agents and entities on your site.
Memory TTL
The Default Memory TTL setting controls how long memories are kept before being automatically deleted. Set the number of days and save.
- Set to 0 to keep memories indefinitely (they will never expire automatically)
- Memories with an explicit TTL set by the agent override this default
Viewing and Filtering Memories
The memory table shows all stored entries with the following columns:
- Type — the category of memory (e.g. preference, decision, context)
- Entity — the user or session the memory belongs to
- Key — the memory identifier
- Value — the stored content (truncated; hover to see the full value)
- Expires — the expiry date, or — if the memory has no TTL
- Updated — when the memory was last written
Use the Type and Entity dropdowns to filter the table. Click Filter to apply.
Deleting Memories
Click the × button on any row to delete that individual memory entry immediately.
To remove multiple entries at once, filter by type or entity and click Clear Filtered to delete all matching entries. Click Clear All Memories to wipe every memory entry on the site.
The memory table is empty until your assistants have had enough conversations to start accumulating context. Once memory entries exist, they appear here automatically.
Frequently Asked Questions
Which agents use memory, and which do not?
Memory is an optional capability. Bundled assistants that benefit from continuity — such as the Content Writer, WordPress Assistant, and customer-facing agents — use memory to persist preferences and decisions between sessions. Assistants focused on one-shot tasks (the Security Assistant running a scan, or the Site Doctor diagnosing a performance issue) typically do not write memory entries because each task is self-contained. Custom agents you build with the Assistant Trainer can be configured to use memory by enabling the memory tools in their tool set.
Is memory shared between different users on the same site?
No. Memory is stored per entity, which is typically the WordPress user ID. Each user’s conversations with an assistant build their own separate memory store. An editor who tells the assistant their preferred writing style does not share that preference with the site administrator. On the Memory tab you can filter by Entity to see which memories belong to which user. If you run public-facing agents (embedded on the front end), memory is stored per session identifier rather than WordPress user ID.
How does the agent decide what to remember?
The agent uses its memory tools autonomously based on its system prompt and the context of the conversation. If a user states a clear preference (“I always want posts in British English”) or a key decision is made (“we agreed to use WooCommerce for the new shop”), the agent flags it as worth storing. You cannot directly instruct the agent to “remember this” in a deterministic way — but agents with memory tools enabled will naturally capture things they determine are relevant for future sessions. You can view and delete any memory entry from the Memory tab if you disagree with what was stored.
Does deleting a memory affect ongoing conversations?
Deleting a memory entry removes it from the database immediately. If the agent has already loaded that memory into its current conversation context, the current session is not affected — it will still “know” what it loaded at the start. The deletion takes effect on the next new conversation. If you want to ensure a deleted preference does not influence the current session, close the chat and start a new one after deleting the entry.
Can I bulk-import or export memory entries?
Not directly from the Memory tab UI — it is designed for viewing and managing individual entries. If you need to migrate memory entries between environments (staging to production, for example), memory is stored in the WordPress database as custom post type entries and can be exported using standard WordPress database tools or a plugin like WP Migrate. For bulk seeding an agent with initial context, use the Persona Notes and Knowledge fields in Agent Instructions instead — those are easier to manage and version control.