🧠 Best way to give long-term memory to an AI Engineer Agent in n8n?

Hey folks!

I’m currently building an AI Engineer Agent inside n8n — basically a smart assistant to help me bring my ideas to life, build other AI agents, keep track of all our past projects and logic, and suggest better solutions based on experience.

Now I’ve hit a bit of a wall when it comes to long-term memory.

The goal is to make this agent actually remember what we’ve done — not just keep everything in short-term context. So it can learn from the past and not waste tokens re-reading everything each time.

Here’s what I’ve tried so far:

I managed to get ChromaDB running on my VPS (yay!). It works, but I’m not yet sure how to structure the memory (projects? prompts? task logs?).
Tried to self-host Zep for 3 days… no luck so far. Couldn’t get the cloud version working either.
I’m now looking into Postgres Chat Memory, Redis, Xata, or even Pinecone (I’ve got an account and n8n node set up already).

👉 So here’s my question to you all:

What would you recommend for storing long-term memory for an AI agent like this?

Something solid, easy enough to manage, and plays nicely with n8n + LLMs.

I’d really appreciate any ideas, experiences, best practices, or examples — even if it’s something simple, I’m happy to learn!

Thanks a ton 🙏

2 comments