[Community Node] n8n-nodes-memori-community Long-term memory for n8n agents via Memori

Tips & Tricks
1

I’ve published a community node that brings Memori long-term memory into n8n workflows.

If you’ve been wanting your AI Agents to actually remember things across sessions, user preferences, prior conversations, durable facts, this plugs straight in.

Note: I am not associated with Memorilabs, this node is unofficial.

:package: npm: https://www.npmjs.com/package/n8n-nodes-memori-community
:hammer_and_wrench: Source: https://github.com/the-automagicians/memori-ai-model

What it does

Memori is an agent-native memory layer that captures structured facts, preferences, relationships, and events from your LLM interactions and recalls the relevant bits automatically on later calls.

In a chatbot the experience is so smooth it’s unreal, fully transprent to n8n, the AI model response has the relevant memories baked in, a cold start session can look like this…

chat-memori-agent
chat-memori-agent969×458 63.4 KB

It’s LLM-agnostic OpenAI, Anthropic, Gemini, Bedrock, DeepSeek, Grok and benchmarks well on long-conversation memory (LoCoMo).

In n8n, the node lets you wire memory into your AI Agent flows.

n8n’s AI Agent does not know what’s going on, it receives an LLM answer with the memories ‘baked in’. For this to work we have to provide three things to Memori

  • Entity - The user id across apps and sessions
  • Process ID - The agent or app the user is interacting with
  • Session ID - The current session

.. so Claude and I wrote a custom AI Model node expanding the built-in openAI LLM module.

This is not a fully free, fully open-source stack:

  • The Memori Proxy with BYODB are open source (Apache 2.0) https://github.com/MemoriLabs/Memori/
    You have to set up the Memori Proxy with FastAPI on a server/VPS somewhere.
    Instructions below. This Community Node will talk to your Memori Proxy, which stores your data in postgres DB and runs the augmentation.

  • This Memori Proxy still requires a Memori API key. The augmentation pipeline runs against Memori’s service. There is a free tier, but it is rate-limited, for any real workload you’ll need a paid subscription.

  • Self-host the Memori proxy and bring your own Postgres (BYODB) if you want to keep your data on your own infrastructure. Docs: https://memorilabs.ai/docs/memori-byodb/

  • So: not “free and open” end-to-end. Closer to “open core + managed service,” with the option to host the data layer yourself.

Install

Settings → Community Nodes → Install → n8n-nodes-memori-community

You’ll need a MEMORI_API_KEY from https://app.memorilabs.ai/

The full source for the node is on GitHub: https://github.com/the-automagicians/memori-ai-model

Self-hosting a Memori proxy for n8n (or any OpenAI-compatible client)

The n8n-nodes-memori-community node speaks OpenAI-compatible HTTP, but the upstream MemoriLabs/Memori project is a Python SDK, not a server. Here’s the minimum to bridge them, a small FastAPI app that wraps the SDK.

  1. Clone the starter gist — five files: main.py, requirements.txt, Dockerfile, docker-compose.yml, .env.example.

  2. Provision a Postgres 14+ database; copy its URL into .env as MEMORI_POSTGRES_URL. Memori auto-creates its schema on first run.

  3. Set the rest of .env: MEMORI_PROXY_API_KEY (any long random string — clients send this as Authorization: Bearer …), OPENAI_API_KEY, and UPSTREAM_BASE_URL if you’re routing through something other than OpenAI direct.

  4. docker compose up -d --build.

  5. curl -s http://localhost:8012/health → {"status":"healthy"}.

  6. In n8n, install n8n-nodes-memori-community, create a Memori API credential with Base URL http://<host>:8012/v1 and the proxy key, and add a Memori Chat Model to your Agent.

1 Like