Resources · n8n integration

n8n Qdrant Vector Store nodeConfigure Qdrant Vector Store in n8n.

A vector store keeps mathematical representations of your content so an AI can look them up by meaning, not by keyword. The n8n Qdrant Vector Store node reads and writes a Qdrant collection through 2 parameters and 4 operation modes. Built for anyone wiring retrieval into a workflow.

Try the Qdrant Vector Store node on n8nSee the 2 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Qdrant Vector Store node actually do?

The node connects a workflow to one Qdrant collection, a named bucket of stored vectors. Depending on the operation mode you pick, it writes documents into that collection, pulls back the documents that rank closest to a search prompt, or exposes the collection to another AI node so a chain or an agent can query it on its own. One node, four jobs, the same collection.

Pattern one: the ingestion side. You set the mode to Insert Documents, feed the node with content coming from earlier steps, and every item lands in the collection. The n8n docs show this shape in a movie recommendation chatbot template (workflow 2440) where documents are split, embedded, then written to Qdrant before anyone asks a question. Nothing answers yet, you are only filling the shelf.

Pattern two: the lookup. Get Many mode takes a Prompt, ranks the stored documents by similarity, and returns the best ones as normal workflow items. That output goes anywhere the rest of n8n goes, into a Slack message, a row in Google Sheets, or a reply drafted through Gmail. This is the mode to reach for when you want the retrieved text in your own hands rather than inside a model prompt.

Pattern three: hand the collection to an AI. Retrieve Documents (As Tool for AI Agent) turns the node into something an AI Agent can call by name, which is why the mode asks for a Name and a Description. Retrieve Documents (As Vector Store for Chain/Tool) does the quieter version: the node sits behind a Vector Store Retriever node, and a Question and Answer Chain asks it for context. The n8n docs document both flows, plus the Vector Store Question Answer Tool variant that summarizes results before the agent sees them.

Where it stops. The node talks to Qdrant collections and nothing else, so anything outside that scope belongs elsewhere in the workflow. Embeddings are produced by a separate node, not this one, which is why templates pair Qdrant with OpenAI or Mistral Cloud embeddings. And the node runs at version 1, so an older workflow may show fewer options than the panel described here. If retrieval quality is the thing you are wrestling with, the n8n training covers the build order rather than the node settings.

Connect

What do you need before the node runs?

  1. 01

    Create the Qdrant credential

    Add the Qdrant credential from the Credentials menu: an API Key and your Qdrant URL. Get the key from the Qdrant Cloud Dashboard, under Access Management, scoped to the cluster you want. Do this once: the credential is stored and reused by every workflow talking to the same instance.

  2. 02

    Attach it to the node

    Drop the Qdrant Vector Store node on the canvas and select that credential in its credential field. The same node powers all 4 operation modes, so the credential you pick here serves the ingestion workflow and the retrieval workflow alike. Nothing else in the panel depends on it, which means you can select the credential before you decide what the node will do.

  3. 03

    Point it at a collection

    Fill Qdrant Collection, the resource locator that names the collection the node reads or writes. It is required in every mode. A resource locator lets you pick from a list or type the value yourself, which matters when the collection name comes from an earlier step as an expression like {{ $json.collection }} instead of being fixed in the workflow.

Parameters

Which parameters do you set in the panel?

The Qdrant 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.

01

Operation Mode

mode

What you see in n8n

Notes & use cases

This selector decides what the node is, and the rest of the panel follows from it.

Key parameters

  • Operation Mode: insert (Insert Documents) writes documents, load (Get Many) returns ranked documents, retrieve (Retrieve Documents (As Vector Store for Chain/Tool)) serves a chain, retrieve-as-tool (Retrieve Documents (As Tool for AI Agent)) serves an agent.
  • Prompt and Limit: the query, and how many top results return, 10 for the ten best.
  • Name and Description: what an agent sees, such as company_knowledge_base and a specific explanation.
  • Include Metadata, Rerank Results, ID: metadata with the text, reordered hits, an embedding entry identifier.
Use cases
a workflow on insert keeps a documentation collection current, another on retrieve-as-tool answers from it.
02

Qdrant Collection

qdrantCollection

What you see in n8n

Notes & use cases

Every mode needs to know which collection it is working against, and this is the field that says so. It is a resource locator, the n8n field type that accepts either a value you select or one you type, and it is marked required.

Key parameters

  • Qdrant Collection: the collection the node reads from or writes to, set once per node, either chosen in the panel or built from an expression such as {{ $json.collection }} when the name arrives with the item.
Use cases
a team that keeps one collection per client picks the value dynamically, so the same retrieval workflow serves every account without a copy per collection.
Need help

Need help automating Qdrant Vector Store with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the Qdrant Vector Store node included in n8n?
Yes. It ships in the n8n AI package, the @n8n/n8n-nodes-langchain bundle that comes with n8n, so there is nothing to install and no extra cost on the n8n side. That holds on n8n Cloud, the hosted offer run by n8n, and on a self-hosted instance under the Community Edition and its Sustainable Use licence. A workflow behaves the same either way. What does get billed sits outside n8n: the provider of the model you use for embeddings charges for its own API calls under its own terms, and n8n adds nothing on top of that.
02What do you need to make the n8n Qdrant Vector Store node work?
A credential and a collection. The credential is created once in the Credentials menu and reused across workflows, and the n8n documentation for this node links to the credentials page that lists the values to provide. Then the Qdrant Collection field, a required resource locator, tells the node which collection to work against. Beyond that it depends on the mode: the search modes want a Prompt, and the agent tool mode wants a Name and a Description so the model knows what it is calling.
03What are the limits of this node?
It works with a Qdrant collection and stops there. It does not produce the embeddings themselves, which is why the published templates pair it with an embeddings node such as OpenAI or Mistral Cloud, along with a text splitter and a data loader. The catalogue lists version 1 for this node, so an older workflow can show fewer options than the current panel. Filtering happens through the Metadata Filter node option, and on Insert Documents the Collection Config option takes JSON for the collection creation configuration.
04What should you connect at a minimum, and where do you start?
Start with the ingestion half. One workflow on Insert Documents, a data loader and a text splitter ahead of it, an embeddings node to turn the text into vectors, and a collection that fills up. Only then add the reading half. The simplest reader is Get Many, because the results come back as plain workflow items you can inspect. Once that returns what you expect, move up to Retrieve Documents (As Tool for AI Agent) and hand the collection to an agent, or place a Vector Store Retriever in front of a Question and Answer Chain.
05n8n or Make for a vector store workflow?
It depends on where the data should live. n8n self-hosts through Docker or npm, or runs on n8n Cloud, so a RAG pipeline pointing at your own Qdrant instance can stay inside your own network end to end. Make is hosted only, with no self-hosting option, and it bills per operation, which is a different cost shape when an ingestion run writes thousands of documents. The visual logic differs too. Pick on hosting, data control, cost model and how you like to build, not on a feature checklist.
Hack'celeration Lab

Get our weekly integration tips.

No spam. Unsubscribe anytime.

Same platform

Other docs you might need on this stack.

Browse all integrations