Long-term AI memory for n8n agents. Drop-in memory layer with knowledge graph, semantic search, entity tracking, and session continuity. Powered by StudioMeyer Memory.
n8n's built-in memory options (Postgres, Redis, Motorhead) handle short-term conversation history but stop there. StudioMeyer Memory is a long-term layer:
- Semantic search across past decisions, learnings, and sessions
- Knowledge graph with typed entities (people, projects, services) and relations
- Bi-temporal tracking so you can query "what did we know about X on date Y?"
- Confidence + decay so old contradicted facts fade automatically
- Multi-tenant on Supabase EU — DSGVO-compliant by default
This community node connects your n8n workflows to a hosted or self-hosted Memory server and exposes the most useful operations as a clean Resource/Operation tree.
A voice agent that remembers callers across sessions. A customer support bot that knows the customer's preferences from three months ago. A personal assistant that builds a profile of you over years, not minutes. n8n is the perfect glue layer — but it lacks a real long-term memory primitive. This node fills that gap.
In your n8n instance: Settings → Community Nodes → Install
Enter the package name:
n8n-nodes-studiomeyer-memory
Or via npm in self-hosted setups:
npm install n8n-nodes-studiomeyer-memory- Get an API key. Sign in at studiomeyer.io/portal/login (Google, GitHub, Discord or email magic link). Inside the portal, click "Free Memory testen" — you get an API key with 200 free credits, no credit card. Copy the key from your dashboard at studiomeyer.io/portal/api.
- Add the credential in n8n. Settings → Credentials → New → "StudioMeyer Memory API". Paste your key.
- Drop the StudioMeyer Memory node into a workflow. Pick a Resource (Memory / Entity / Session / Insight) and an Operation.
The four core read/write operations.
| Operation | Maps to | Use case |
|---|---|---|
| Search | nex_search |
Semantic + keyword search across all memory types with temporal decay |
| Recall | nex_recall |
Faster lightweight recall (full-text only) |
| Learn | nex_learn |
Persist a fact, pattern, mistake, or workflow note |
| Decide | nex_decide |
Record a decision with rationale and confidence |
Knowledge graph operations.
| Operation | Maps to | Use case |
|---|---|---|
| Create | nex_entity_create |
Add a person, project, service, or concept |
| Observe | nex_entity_observe |
Append observations to an existing entity |
| Search | nex_entity_search |
Fuzzy search across names + observations |
| Relate | nex_entity_relate |
Connect entities (uses, depends_on, competes_with, custom) |
| Open | nex_entity_open |
Fetch one entity with all observations + relations |
Manage long-running agent sessions.
| Operation | Maps to | Use case |
|---|---|---|
| Start | nex_session_start |
Open a session, load context, get a session ID |
| End | nex_session_end |
Close a session and persist a summary |
| Recall Timeline | nex_recall_timeline |
Chronological view of recent activity |
Higher-level synthesis.
| Operation | Maps to | Use case |
|---|---|---|
| Synthesize | nex_synthesize |
Cluster recent learnings into an insight |
| Reflect | nex_reflect |
Surface emerging patterns + contradictions |
| Proactive Briefing | nex_proactive |
Status briefing — stale learnings, open decisions, knowledge gaps |
Two modes, same endpoint:
- API Key (default, recommended). Paste a key from your portal at studiomeyer.io/portal/api. First-time users sign in once at /portal/login, click "Free Memory testen", and the API key appears in the dashboard. 200 credits, no card.
- OAuth 2.1 Access Token. Pre-issued token from a PKCE flow against
memory.studiomeyer.io/authorize+/token. The OAuth discovery doc is at memory.studiomeyer.io/.well-known/oauth-authorization-server. The fully integrated browser flow (n8n OAuth2 credential type) ships in v0.2.
Both modes target https://memory.studiomeyer.io/mcp. Set a custom Server URL in the credential for self-hosted deployments.
Voice Webhook (Vapi/Retell)
→ Memory: Search (query: caller_phone, types: ["entity", "learning"])
→ AI Agent (with caller context in system prompt)
→ Memory: Learn (content: "Caller X asked about Y, agent answered Z")
WhatsApp Trigger
→ Entity: Search (query: customer_email, entityType: customer)
→ Entity: Open (entityRef: matched_id)
→ Anthropic Claude (with full customer context)
→ Memory: Learn (tag the ticket outcome)
Telegram Trigger
→ Session: Start (project: telegram-bot, agentId: user_id)
→ Memory: Search (query: telegram_message, recencyWeight: 0.5)
→ AI Agent
→ Memory: Learn (whatever the user just told you)
Free tier (1000 learnings + 100 entities) is enough for a single bot in development. Pro is €29/month and lifts the limits. Team is €49/month with multi-agent isolation. See studiomeyer.io/services/memory for the latest pricing.
Self-hosted is free under MIT for the client; the server is source-available with the SaaS-multi-tenant pieces under a separate license. Contact for self-hosted licensing details.
- v0.1 (current): Memory / Entity / Session / Insight resources, API Key auth, OAuth 2.1 access token (manual). v0.1.1 ships doc fixes (correct dashboard URL, clarified OAuth setup).
- v0.2: AI-Agent Memory Sub-Node (drops directly into the n8n AI Agent's Memory slot, replacing Postgres / Redis / Motorhead). Dedicated n8n OAuth2 credential type with full browser-based PKCE flow.
- v0.3: Streaming support for long search results, batch operations
- v0.4: Workflow templates published to n8n.io marketplace
git clone https://github.com/studiomeyer-io/n8n-nodes-studiomeyer-memory
cd n8n-nodes-studiomeyer-memory
npm install
npm run build
npm testTo test locally in your n8n instance:
npm link
cd ~/.n8n/custom
npm link n8n-nodes-studiomeyer-memory
n8n startPull requests welcome. Please open an issue first for non-trivial changes.
MIT — see LICENSE.