Description

n8n-nodes-memorysync

MemorySync nodes for n8n — long-term
memory for your workflows and a persistent brain for n8n’s AI Agent.

Two nodes in one package:

| Node | What it does |
| — | — |
| MemorySync | Workflow operations: Add Memory, Add Conversation Turn, Search, Recall Context, Get Many, Delete. Marked usableAsTool, so an AI Agent can call it autonomously. |
| MemorySync Chat Memory | An AI Agent memory sub-node (AiMemory output): the agent’s conversation history persists in MemorySync across executions — sessions survive restarts, redeploys, and weeks between chats. |

Install (self-hosted n8n)

Settings → Community Nodes → Install → n8n-nodes-memorysync

(n8n Cloud surfaces only verified community nodes; verification submission is in progress.)

Credentials

Create a MemorySync API credential with your API key from
app.memorysync.io → Settings → API Keys. The built-in credential test makes a
real (tiny) query, so a bad key fails at setup time — not mid-workflow.

The AI Agent memory sub-node

Chat Trigger ──▶ AI Agent ◀──memory── MemorySync Chat Memory
                        ◀──tool────── MemorySync (Search / Add)

• Session ID (e.g. {{ $json.sessionId }}) scopes the transcript; the same

session resumes its history on the next execution.

• End User ID isolates everything per user.
• Context Window Length counts turn pairs handed to the agent each run.
• Turns store verbatim with deterministic idempotency seeds — a retried

execution can never duplicate a conversation.

• clear() is a deliberate no-op: an agent can never bulk-delete a

customer’s stored history. Deletion stays an explicit human action.

• A MemorySync outage degrades to “no history this run” — the agent still

answers; nothing throws.

Requires an n8n version that ships @n8n/ai-node-sdk (current releases do).
The regular MemorySync node works on any n8n with community-node support.

Operations cheat-sheet

| Operation | Use it for |
| — | — |
| Add Memory | “Remember this fact about the customer” (server-side extraction applies) |
| Add Conversation Turn | Verbatim transcript capture with duplicate-proof retries |
| Search Memories | Semantic lookup — results fan out as items |
| Recall Context | A grouped, prompt-ready context block for a downstream AI step (unique to MemorySync) |
| Get Many | Newest-first listing for a user |
| Delete Memories | Explicit deletion by ID (m_123) |

Semantics worth knowing

• Every operation is scoped by End User ID — header-enforced isolation.
• Free-tier quota exhaustion is silent by design: adds answer accepted-without-

storing, reads answer empty. The node reports stored: false, accepted: true
instead of failing your workflow. Evaluation keys surface strict 429s.

• Transcript turns store under the n8n:: scope — separate history,

same shared user memories as every other MemorySync surface.

Development

npm install
npm test        # builds + runs 23 contract/behaviour tests (node --test)

License

MIT