Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.memanto.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
n8n + Memanto

n8n Add persistent memory to your n8n workflows using Memanto’s REST API. n8n workflows are stateless by default — each execution starts fresh. With Memanto, you can store context from one workflow run and recall it in future runs, giving your automations a memory that grows over time.

​
How It Works

n8n Workflow -> HTTP Request nodes -> Memanto Server -> Moorcheh.ai
Memanto exposes a simple REST API. In n8n, you call it using the built-in HTTP Request node — no custom code or plugins required.
Memanto authenticates with Moorcheh on the server using MOORCHEH_API_KEY. n8n nodes do not send an Authorization header. The only header you pass from n8n is X-Session-Token for memory operations.

​
Prerequisites

  • n8n (self-hosted or cloud)
  • Memanto server accessible from your n8n instance, with MOORCHEH_API_KEY configured in its environment

​
Step 1: Start Memanto Server

On your server or locally: 
export MOORCHEH_API_KEY=your_moorcheh_key
pip install memanto
memanto serve
If n8n is running in the cloud or Docker, expose Memanto via a public URL or use a tunnel like ngrok.

​
Core Workflow Patterns

​
Pattern 1: Activate Session + Remember

Use this at the start of a workflow to open a session and store context. Node 1 — Activate Session (HTTP Request)
FieldValue
MethodPOST
URLhttp://your-memanto-server:8000/api/v2/agents/n8n-agent/activate
AuthenticationNone
This returns a session_token. Reference it in later nodes via: 
{{ $node["Activate Session"].json.session_token }}
Node 2 — Remember (HTTP Request)
FieldValue
MethodPOST
URLhttp://your-memanto-server:8000/api/v2/agents/n8n-agent/remember
HeadersX-Session-Token: {{ $node["Activate Session"].json.session_token }}
JSON Body{ "content": "{{ $json.message }}", "type": "fact" }

​
Pattern 2: Recall Context

Retrieve relevant memories before making an LLM call or sending a response. Node — Recall (HTTP Request)
FieldValue
MethodPOST
URLhttp://your-memanto-server:8000/api/v2/agents/n8n-agent/recall
HeadersX-Session-Token: {{ $node["Activate Session"].json.session_token }}
JSON Body{ "query": "{{ $json.userMessage }}", "limit": 5 }
The response contains a memories array. Access the first result with: 
{{ $json.memories[0].content }}
Or join all results into a single string using a Code node: 
const memories = $input.first().json.memories || [];
return [{ json: { context: memories.map(m => `- ${m.content}`).join("\n") } }];

​
Pattern 3: AI-Powered Answer from Memory

Let Memanto answer a question directly using its built-in RAG: Node — Answer (HTTP Request)
FieldValue
MethodPOST
URLhttp://your-memanto-server:8000/api/v2/agents/n8n-agent/answer
HeadersX-Session-Token: {{ $node["Activate Session"].json.session_token }}
JSON Body{ "question": "{{ $json.question }}" }
Returns answer — a grounded response based on stored memories, no external LLM call needed.

​
Example: Customer Support Workflow

This workflow receives a customer message via webhook, recalls past context, and sends a personalized reply. 
[Webhook] → [Activate Session] → [Recall Context] → [Format Prompt] → [OpenAI] → [Remember Exchange] → [Respond]
Webhook Node — receives: 
{ "customer_id": "cust_123", "message": "I need help with my order" }
Activate Session — POST /api/v2/agents/{{ $json.customer_id }}/activate Using the customer ID as the agent ID gives each customer their own isolated memory. Recall Context — POST /api/v2/agents/{{ $json.customer_id }}/recall with body { "query": "{{ $json.message }}", "limit": 5 } Code Node — Format Prompt 
const memories = $input.first().json.memories || [];
const context = memories.length
  ? memories.map(m => `- ${m.content}`).join("\n")
  : "No prior context.";

return [{
  json: {
    prompt: `Customer history:\n${context}\n\nCustomer: ${$("Webhook").item.json.message}\nAgent:`
  }
}];
OpenAI Node — use the formatted prompt to generate a reply. Remember Exchange — POST /api/v2/agents/.../remember with body { "content": "<message + reply>", "type": "fact" }

​
Memory Types in n8n

Set the type field in the Remember body to categorize what you store:
TypeWhen to use
factObjective information about the user or entity
preferenceUser likes, dislikes, or settings
decisionChoices made during the workflow
commitmentPromises or follow-up actions
eventThings that happened (order placed, ticket opened)
errorIssues or failures to avoid repeating

​
Sessions in Long-Running Workflows

Memanto auto-renews sessions that are near expiry on memory requests, so most workflows can simply reuse the token from the activation step. There is no separate /session/extend endpoint. For scheduled or long-lived workflows, the simplest pattern is to activate a fresh session at the start of each run rather than persisting the token between executions: 
[Trigger] → [Activate Session] → ...remaining nodes use the new token...
If a request returns 401 Unauthorized, re-run the Activate Session node and continue with the new token.

​
Next Steps