Description

n8n-nodes-memory

Custom memory nodes for n8n AI Agent with workflow-based storage.

Two Nodes

| Node | Use Case |
|——|———-|
| Memory API | External webhook/API as memory backend |
| Memory Workflow | Sub-workflow within n8n as memory backend |

The Power: Memory as Workflows

Traditional memory nodes are passive storage. This node transforms memory into event-driven workflows – each memory operation (get, add, clear) can trigger complete n8n workflows with access to 400+ integrations.

On GET (before AI responds)

• Fetch context from vector databases (RAG)
• Load user preferences from CRM
• Inject real-time data (weather, stocks, news)
• Summarize old conversations automatically
• Generate dynamic context with another AI
• Apply predictive loading for likely questions

On ADD (after each message)

• Analyze sentiment in real-time
• Extract entities and intents
• Update user profiles and CRM records
• Create tasks/tickets automatically
• Trigger notifications and alerts
• Route to different AI agents based on intent
• Build knowledge graphs from conversations
• Feed training data pipelines

On CLEAR (end of session)

• Generate conversation summaries
• Extract action items
• Send follow-up emails
• Archive to compliance systems
• Update customer journey stages

Use Cases

Self-Evolving AI – Behavior rules that adapt based on conversation patterns. Detect frustration, switch to empathetic mode.

Multi-Agent Orchestration – Route messages to specialized agents. One memory feeding an agent swarm.

Memory as API Gateway – Natural language interface to any system. “Check my order status” triggers lookup workflows.

Semantic Compression – Compress long exchanges into dense summaries, expand on retrieval. Effectively infinite context window.

Cross-Platform Identity – Sync across WhatsApp, Telegram, Web. Same AI remembers you everywhere.

Conversation Branching – Create save points, fork conversations, explore different paths. Git for chat.

Regulatory Firewall – Check compliance before storing, redact sensitive info based on user role. GDPR/LGPD by design.

Continuous Learning – Route high-quality exchanges to training datasets. Self-improving system.

Social Graph Memory – Map relationships mentioned in conversations. Build knowledge graphs of user’s world.

Emotional State Machine – Track emotional journey, maintain consistent AI “mood” across sessions.

Installation

npm install n8n-nodes-memory

Then restart n8n.

Setup

1. Create a Memory API workflow – A webhook that handles get, add, and clear actions
2. Set up your storage – PostgreSQL, Redis, or any backend
3. Connect to AI Agent – Link the Memory API node to your agent’s memory input

API Contract

Your webhook must handle POST requests with:

| Action | Purpose | Returns |
|——–|———|———|
| get | Retrieve messages for session | { messages: [{ type, content }, ...] } |
| add | Store a message | { success: true } |
| clear | Clear session history | { success: true } |

Message types: human or ai

Parameters

Memory API

| Parameter | Description |
|———–|————-|
| API URL | Your memory webhook URL |
| Session ID | Unique conversation identifier |
| API Key | Optional Bearer token |
| Context Window Length | Messages to include (default: 10) |

Memory Workflow

| Parameter | Description |
|———–|————-|
| Workflow ID | ID of the sub-workflow to execute |
| Session ID | Unique conversation identifier |
| Context Window Length | Messages to include (default: 10) |

How It Works

Implements LangChain’s BaseListChatMessageHistory with BufferWindowMemory. The AI Agent calls your workflows automatically during conversation.

License

MIT

Author

Filipe Labs