n8n Zep Vector Store nodeConfigure Zep Vector Store in n8n.
Zep stores your documents as vectors, and this node is how a workflow puts them in and gets them back out again. The n8n Zep Vector Store node ships with 2 parameters and 4 operation modes, from a plain insert to a tool an AI agent can query on its own.
Verified Trustpilot reviews · AI, automation & growth agency
What does the n8n Zep Vector Store node actually do?
The node connects a workflow to a Zep vector database, the store that keeps your text as arrays of numbers so a search can match meaning instead of exact words. One parameter picks what it does, Operation Mode, and another says where, Collection Name. Everything else on the panel appears or disappears depending on the mode you chose.
Start with insert (Insert Documents). A workflow reads a batch of support articles, splits them, and the node writes them into a named collection. Nothing is queried yet, you are just filling the store. The Is Auto Embedded option stays on by default here, which means Zep builds the embeddings itself instead of n8n doing it upstream.
Then comes the reading side, and this is where the four modes stop looking alike. load (Get Many) runs a single search and hands the matching documents back to the next node, as workflow items you can mail, log or post. retrieve (Retrieve Documents (As Vector Store for Chain/Tool)) does not return anything by itself, it exposes the store so a chain can pull from it, a chain being a fixed sequence of calls to a model. retrieve-as-tool turns the store into a tool an AI Agent decides to call when a question needs your data, with Name and Description telling the model what it is looking at.
The official docs describe four connection patterns around the node. It can sit in the regular flow to insert and get documents. It can be wired straight to the tool connector of an AI agent. A Vector Store Retriever node can read from it and feed a Question and Answer Chain. Or a Vector Store Question Answer Tool can sit between the agent and the store to summarize results instead of returning raw documents.
Now the part that decides everything else: the Zep Vector Store node is deprecated from n8n 1.108.0. If you are picking a store for something new, pick another one. This page is for the workflows you already run and the instances still below that version. The Zep team also deprecated the open source Zep Community Edition in April 2025, so the self-hosted setup path may no longer behave as documented.
Model calls made around the node are billed by whoever provides the model, OpenAI or another provider, under their own terms. n8n adds nothing on top. If you want the wider picture before committing a retrieval stack, the n8n review covers where the platform fits, and the n8n training goes through building these workflows hands on.
How do you connect Zep to n8n?
- 01
Create the key in Zep
You need a Zep server with at least one project. In Zep Cloud, open Project Settings, go to the Project Keys section and select Add Key. Give it a Key Name that says where it will be used, something like
n8n integration, then select Create. Copy the key right away, that screen is the one chance you get to read it. - 02
Save the credential in n8n
In n8n, open the Credentials menu and create a Zep credential, the stored login n8n reuses across workflows so no key ever sits inside a node. The supported authentication method is an API key, and the credential holds two fields: an API URL and an API Key. Paste the key you copied into API Key, then turn the Cloud toggle on.
- 03
Or point it at your own server
Self-hosting Zep Open Source changes all three values. Enter the JWT token of your Zep server as the API Key, leave the Cloud toggle off, and put the URL of your server in API URL. Same credential, different meaning for each field. The same Zep credential also authenticates the Zep memory node, so you create it once and select it in both.
Which parameters does the Zep Vector Store node expose?
The Zep Vector Store node has 2 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.
Operation Mode
modeWhat you see in n8n
Notes & use cases
This selector sets the role of the node: writer, reader, store for a chain, or agent tool. Fix it first.
Key parameters
- Operation Mode:
insert(Insert Documents),load(Get Many),retrieve(Retrieve Documents (As Vector Store for Chain/Tool)),retrieve-as-tool(Retrieve Documents (As Tool for AI Agent)). - Prompt and Limit: the search query, often
{{ $json.question }}, and how many top results come back, say10. - Name and Description: what the model reads before calling the store.
- Include Metadata, Rerank Results and ID: metadata, reordered matches, the identifier of an embedding entry.
insert fills the store overnight, another on retrieve-as-tool answers from it by day.Collection Name
collectionNameWhat you see in n8n
Notes & use cases
The collection is the named bucket inside Zep where the vectors live. This field says which one the node writes to or reads from, and it is required in every mode, so a typo here is the difference between a search that returns nothing and one that works.
Key parameters
- Collection Name: a required string. Fixed text when a workflow always serves one dataset, or an expression like
{{ $json.workspace }}when one workflow serves several.
Need help automating Zep Vector Store with n8n?
A person reads every message.
Zep Vector Store and n8n, the questions that come up
01Is the Zep Vector Store node included in n8n?
02What do you need to make the n8n Zep Vector Store node work?
03What are the limits of this node?
04What do you plug the node into to get started?
05n8n or Make for this kind of workflow?
Get our weekly integration tips.
No spam. Unsubscribe anytime.
