Resources · n8n integration

n8n Vector Store Question Answer Tool nodeConfigure Vector Store Question Answer Tool in n8n.

An agent that reads an entire knowledge base before every answer is an agent nobody waits for. The n8n Vector Store Question Answer Tool node searches an indexed store, summarizes what comes back and hands the agent a finished answer. Three parameters, two ports to fill, and it runs.

Try the Vector Store Question Answer Tool node on n8n See the 3 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Vector Store Question Answer Tool node do?

It is a tool, meaning an add-on the AI model can call when it needs something it does not know. Attached to an agent, it takes a fully formed question, searches the vector store wired to it, and returns a summarized answer built from the matching chunks. The agent never sees the raw documents. It asks, and it gets a paragraph back.

A vector store keeps mathematical representations of your text so that a search returns passages by meaning rather than by exact wording. This node sits on top of one. It is a sub-node, a helper block that never runs on its own: it only executes when a root node such as an AI Agent decides to call it.

Take a support chatbot over company documents. Files land from Google Drive, a Recursive Character Text Splitter cuts them into chunks, Embeddings Google Gemini indexes them in a Pinecone Vector Store, and this tool answers staff questions through a Google Gemini Chat Model. That is template 2753 on n8n.io, start to finish.

Second case, a WhatsApp assistant. Template 2465 pairs WhatsApp Business Cloud with an AI Agent, a Simple Memory so the conversation holds together, a Simple Vector Store for the product documentation, and this tool with an OpenAI Chat Model behind it. The customer types a question, the agent routes it here, the answer comes back in the thread.

Third case, a voice and phone agent. Template 3657 indexes documents in a Qdrant Vector Store, adds Google Calendar for bookings, and gives the agent this tool so it can quote the documentation while it talks. Same node, different front door.

When would you skip it? If the model already knows enough to answer and there is no document set behind the question, a Chat Model on its own is enough, which is what a Basic LLM Chain does in that same template. And if your agent needs to act on a system rather than read about it, the tool you want is an application node exposed to the agent, not this one. Check the ports first: the store and the model are both required, so a half-wired node blocks the run rather than degrading quietly. The scope is narrow on purpose, this node answers questions and does nothing else, no writing to the store, no file ingestion. Version 1 is what the catalog carries today. If you want the wider picture on where n8n fits against other platforms, the n8n review covers it, and n8n training goes through agent building step by step.

Connections

What do you plug into it?

n8n splits AI work between a root node that receives the workflow items and sub-nodes that hang off its ports, one type of sub-node per port. Here are the three ports on this node.

Ports (what plugs in)

Vector Storeai_vectorStorerequired
Chat Modelai_languageModelrequired

Output (what it plugs into)

Toolai_tool

01Vector StoreThe store that holds your indexed chunks. This is where the question actually gets searched, so nothing useful comes out of the tool until something has been indexed here first.

The store that holds your indexed chunks. This is where the question actually gets searched, so nothing useful comes out of the tool until something has been indexed here first.

What you drag onto this port is the store sub-node, the same one that indexed the documents in the first place. It decides where the vectors physically live, and the tool only searches what is already sitting there.

Key parameters

Required: leave this port empty and the node refuses to run.

Use cases

a knowledge base indexed once and queried by several agents, each one pointing its own tool node at the same store.

02Chat ModelThe model that turns the retrieved chunks into a written answer. The search finds passages, this model reads them and composes the reply, so it is doing summarizing work rather than raw generation.

The model that turns the retrieved chunks into a written answer. The search finds passages, this model reads them and composes the reply, so it is doing summarizing work rather than raw generation.

Key parameters

Required: the node needs a model of its own, even when the agent above it already has one.
OpenAI Chat Model: the one wired into the WhatsApp template.
Anthropic Chat Model: another provider, same port, same wiring.
Google Gemini Chat Model: paired with Gemini embeddings in template 2753.

Use cases

a lighter model here than on the agent, since the job is condensing retrieved text rather than planning the conversation.

03ToolThe output side. This node does not connect into the main flow, it attaches upward to a root node that is allowed to call tools.

The output side. This node does not connect into the main flow, it attaches upward to a root node that is allowed to call tools.

Key parameters

Required: a sub-node never runs on its own, so without a root node above it nothing ever calls the tool.
AI Agent: the usual destination, in all three published templates.
AI Agent Tool: an agent itself exposed as a tool to another agent.
MCP Server Trigger: publishes the tool to a client outside n8n.

Use cases

exposing an indexed handbook through AI Agent Tool so a second agent can query it without owning the store.

Parameters

Which parameters matter?

The Vector Store Question Answer Tool node has 3 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

Data Name

name

What you see in n8n

Notes & use cases

Names the body of data the tool searches. n8n drops that name into the tool description the model reads before deciding whether to call the tool: "Useful for when you need to answer questions about [name]." Spaces are converted to underscores on the way through.

Key parameters

Data Name: a string, placeholder e.g. users_info. Stick to letters, digits, spaces, dashes and underscores; a special character makes the agent run fail on an invalid function name.

Use cases

one agent holding two stores, one named hr_handbook and one named product_docs, so the model picks the right one from the name alone.

Description of Data

description

What you see in n8n

Notes & use cases

Describes what is inside the store, in plain words. It fills the second half of the generated tool description: "Whenever you need information about [data description], you should ALWAYS use this." Vague wording here is the usual reason an agent ignores a tool that would have answered the question.

Key parameters

Description of Data: a string, placeholder [Describe your data here, e.g. a user's name, email, etc.]. Write it for the model, not for a colleague.

Use cases

"shipping policies, return windows and warranty terms for the online store" tells the agent exactly when a customer question belongs here.

Limit

topK

What you see in n8n

Notes & use cases

Caps how many results come back from the search before the model summarizes them. Fewer results keep the answer tight and the call cheap; more results cover a question whose answer is scattered across several passages.

Key parameters

Limit: a number, the maximum number of results to return. It can take an expression such as {{ $json.topK }} when the value comes from the item feeding the workflow.

Use cases

a policy question answered from one clause runs fine at a low limit, while "summarize everything we say about returns" needs a higher one.

Need help

Need help automating Vector Store Question Answer Tool with n8n?

A person reads every message.

FAQ

Vector Store Question Answer Tool and n8n, the questions people ask

01Is the Vector Store Question Answer Tool free in n8n?

Yes on the n8n side. The node ships inside n8n's AI package, so there is nothing to install and nothing extra to pay, on n8n Cloud and on a self-hosted instance under the Community Edition and its Sustainable Use license. A workflow behaves the same either way. What does cost money sits one port away: the chat model you attach comes from a provider such as OpenAI, Anthropic, Google or Mistral, and that provider bills its own API calls under its own terms. n8n adds nothing on top of that. The vector store is the same story, whatever you index into is billed by whoever runs it.

02What credentials does the Vector Store Question Answer Tool need?

None. This node has no credential and no Authentication selector, so there is nothing to configure on the account side and no key to paste. The authentication lives in the sub-nodes you attach to it. The Chat Model on the ai_languageModel port carries the provider credential, and the vector store sub-node carries whatever its own service needs. That is worth knowing when you are debugging: an authentication error raised here always comes from one of the attached sub-nodes, never from the tool itself, so open the model or the store before you touch anything on this node.

03What are the limits of the n8n Vector Store Question Answer Tool node?

It reads and it answers, nothing else. There is no operation for writing into the store, no file ingestion, no index management; that work belongs to the vector store node and the loaders and splitters around it. It also never runs on its own. As a sub-node it only executes when a root node calls it, so a workflow without an agent above it has nothing to fire it. Both input ports are required, the store and the chat model, and an empty one blocks the execution. The node is at version 1 in the catalog, and an older workflow may show fewer options than this page describes.

04What has to be attached at minimum, and which sub-node is the easiest start?

Two things, both required: a vector store on the ai_vectorStore port and a chat model on the ai_languageModel port. Skip either one and the node will not run. To get moving quickly, a Simple Vector Store is the shortest path because it needs no external service, which is how template 2465 builds its WhatsApp chatbot. Pair it with an OpenAI Chat Model or a Google Gemini Chat Model, whichever provider you already hold a key for. Then attach the whole thing to an AI Agent, fill in Data Name and Description of Data, and the agent starts calling it on its own.

05n8n or Make for a retrieval agent like this?

It depends on where you want the workflow to live. Make runs hosted only and bills per operation, which suits teams that want no infrastructure and predictable clicking. n8n runs on Make's terms too through n8n Cloud, but it can also run on your own Docker or npm install, which matters when the indexed documents are sensitive and you would rather they never leave your network. The retrieval building blocks differ as well: n8n exposes the store, the model and the tool as separate sub-nodes you wire yourself, so you see and control each hop. That is more setup and more room to tune.

Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.