Memory Tools

Memory is Zelaxy's built-in persistent storage for agent conversation history and other keyed data. Use these tools when a workflow needs to remember information across runs — for example, accumulating chat turns under a record identifier, reading back history to feed an agent, or clearing stale records.

This is an internal Zelaxy store (it calls the platform's own /api/memory endpoint), so there is no external service to sign up for and no API key to configure. Records are automatically scoped to the workflow that writes them via the execution context.

Overview

Operations

Configuration

memory_add

Stores a new record or appends to an existing record that shares the same identifier. Every record is stored as agent type, and each entry is a { role, content } message object.

No secrets are required. The workflow identifier is supplied automatically from the execution context (not a user-provided parameter).

memory_get

Retrieves a single record by its identifier.

memory_get_all

Retrieves all records for the current workflow, with optional filtering and sorting.

memory_delete

Permanently removes a record.

Outputs

memory_add

• success (boolean) — Whether the record was stored successfully.
• memories (array) — Array of record (message) objects including the new or updated data.
• count (number) — Number of records in the updated storage.
• error (string) — Error details if the storage operation failed.

memory_get

• success (boolean) — Whether the record was retrieved successfully.
• memories (array) — Array of record data for the requested identifier, with the applied limit.
• count (number) — Number of memory entries retrieved within the specified limit.
• totalCount (number) — Total number of memory entries available in the record.
• limit (number) — The retrieval limit that was applied.
• sortOrder (string) — The sort order that was applied (asc/desc).
• message (string) — Operation result message with limit and count information.
• error (string) — Error details if retrieval failed.

memory_get_all

• success (boolean) — Whether all records were retrieved successfully.
• memories (array) — Array of memory records with metadata and timestamps (key, type, data, createdAt, updatedAt).
• count (number) — Number of records returned after filtering.
• total (number) — Total number of records before filtering.
• message (string) — Operation result message.
• error (string) — Error details if the operation failed.

memory_delete

• success (boolean) — Whether the record was removed successfully.
• message (string) — Operation result confirmation.
• error (string) — Error details if removal failed.

YAML Example

memory_1:
  type: memory
  name: "Memory Storage"
  inputs:
    operation: "add"
    id: "conversation-{{start.input.sessionId}}"
    role: "user"
    content: "{{start.input.message}}"
  connections:
    outgoing:
      - target: agent-1

# Later in the workflow, read the history back:
memory_2:
  type: memory
  name: "Memory Storage"
  inputs:
    operation: "get"
    id: "conversation-{{start.input.sessionId}}"
    limit: 10
    sortOrder: "desc"
  connections:
    outgoing:
      - target: agent-1

Tips

• No API key is needed. Memory is an internal Zelaxy store, and records are automatically scoped to the workflow that creates them via the execution context — there is no apiKey/token parameter to set.
• Reuse the same id to build up a conversation. memory_add appends to an existing record with that identifier instead of overwriting it, so a stable per-session id (for example conversation-{{start.input.sessionId}}) accumulates chat turns over time.
• For memory_get and memory_get_all, use sortOrder: "desc" with a small limit to pull only the most recent turns into an agent's context, and filterType (on getAll) to read back just user, assistant, or system messages.
• The role field only accepts user, assistant, or system; all records are stored internally as the agent type. Reference outputs in downstream blocks with double braces, e.g. {{memory_2.memories}}.