Memory

The Memory block stores and retrieves conversation messages that persist across workflow executions. It is designed to work alongside Agent blocks, giving your agents long-term conversational context that survives between runs.

Overview

Add Memory: Store a message with a role (user, assistant, or system) under a named memory ID

Get Memory: Retrieve a single memory entry by its ID

Get All Memories: Fetch every stored memory for the current workflow

Delete Memory: Remove a memory entry by ID

How It Works

Memory entries are stored as chat-style messages with three required fields: an ID to scope the memory, a Role (user, assistant, or system), and the Content text. When you later retrieve memories and pass them into an Agent block's Memories input, the agent sees them as prior conversation turns -- enabling multi-turn dialogue across separate executions.

A typical conversational workflow uses three Memory blocks around an Agent:

Memory (Add, role: user) -- save the incoming user message before the Agent runs
Agent -- processes the message with retrieved memories as context
Memory (Add, role: assistant) -- save the Agent's response after it runs

Memory IDs scope the conversation. Use the same ID across executions to build a continuous conversation history, or use different IDs to maintain separate threads.

Operations

Stores a new message in the memory store.

Field	Required	Description
ID	Yes	Memory identifier -- groups messages into a conversation
Role	Yes	`user`, `assistant`, or `system`
Content	Yes	The message text to store

Retrieves all messages stored under a specific memory ID.

Field	Required	Description
ID	Yes	The memory identifier to retrieve

Returns every memory entry for the current workflow. No additional fields required.

Removes a memory entry by ID.

Field	Required	Description
ID	Yes	The memory identifier to delete

Configuration

Operation

Select operation

ID*

Enter memory identifier

Role*

Select agent role

Content*

Enter message content

Outputs

Output	Type	Description
`memories`	json	The retrieved memory data (messages array)
`id`	string	The memory identifier that was operated on

Best Practices

Use consistent IDs -- the memory ID is the key that groups messages into a conversation. Use a stable identifier (such as a user ID or session ID) so memories accumulate correctly across runs.
Store both sides of the conversation -- add a user role message before the Agent and an assistant role message after it so the full dialogue is preserved.
Clean up when needed -- use the Delete operation to clear stale conversations and prevent unbounded memory growth.