Persistent Memory for n8n AI Agents - Unison Integration

Learn how to give n8n AI agents persistent team memory using two HTTP Request nodes - no custom node required.

Recall (before your AI Agent node)

HTTP Request node:

• Method GET, URL https://brain.unisonlabs.ai/v1/brain/context
• Query parameter q: your task / user message expression
• Header Authorization: Bearer usk_live_... (store the key in an n8n credential)

Map contextMd from the response into the agent node's system message (skip it when weakEvidence is true).

Write back (after the agent node)

HTTP Request node:

• Method POST, URL https://brain.unisonlabs.ai/v1/brain/ingest
• JSON body:

{
  "items": [{
    "type": "conversation",
    "sourceRef": "{{ $workflow.id }}-{{ $execution.id }}",
    "turns": [
      { "role": "user", "content": "{{ $json.input }}" },
      { "role": "assistant", "content": "{{ $json.output }}" }
    ]
  }]
}

Extraction runs async server-side; the next workflow execution's recall already knows what this one learned.