Resources · n8n integration

n8n Workflow Retriever nodeConfigure Workflow Retriever in n8n.

Sometimes the answer your AI chain needs lives in an API call, not in a vector database. The n8n Workflow Retriever node hands the lookup to another workflow and returns what that workflow produces. It exposes 8 parameters, has no main input, and plugs into a root node through a single retriever port.

Try the Workflow Retriever node on n8nSee the 8 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n Workflow Retriever node do?

The Workflow Retriever node lets a Question and Answer Chain pull its source documents from an n8n workflow instead of a stored index. When the chain needs context, this node calls the workflow you point it at, passes along the values you defined, and gives the returned items back to the chain as retrieved content. It is a sub-node, meaning it never runs on its own: it sits under a root node and only wakes up when that node asks for data.

Think of the three shapes this takes in practice. First shape: internal search that already exists. A support team has a workflow that queries a ticketing API and returns matching threads. Pointing Source at database and entering that workflow's ID turns the existing search into the retrieval step of a question and answer chain, with no index to build and nothing to re-embed when a ticket changes.

Second shape: data that is too fresh to index. Stock levels, open invoices, today's bookings. A workflow that reads a spreadsheet and filters the rows can be called at question time, so the chain always sees the current state. This is where the node pairs naturally with a Google Sheets workflow, or with anything you already automate through Slack or Gmail.

Third shape: routing. Because the called workflow receives the values you set through the workflow values fields, you can send it a category, a customer identifier or a language code and let it decide which source to query. The retrieval logic lives in a workflow you can test on its own, which is easier to debug than a prompt.

When would you skip it? If your documents are static files and the questions are semantic, a vector store retriever fits better: it was built for similarity search and does not pay the cost of a workflow execution per question. If you want the model itself to decide when to fetch something, a tool on an AI Agent gives that control instead, since a retriever is called by the chain, not chosen by the model. And if you simply want a chat model to answer from the prompt, connect OpenAI or Anthropic and leave retrieval out of the design.

The known limits are worth stating up front. This node has no main input and no settings for how many documents to return, how to score them or how to cache them: whatever the called workflow returns is what the chain receives, so filtering and limiting belong inside that workflow. The page describes version 1 of the node, the highest one in the catalog; an older workflow may show fewer options. Nothing here is billed by n8n, though the provider behind whichever chat model you connect charges for its own API calls.

Connections

Where does this node plug in?

n8n splits AI work between a root node, which receives the workflow items, and sub-nodes that attach to its ports, one type of sub-node per port. This node is a sub-node, so the only connection that matters is the port it feeds.

Output (what it plugs into)

  • Retrieverai_retriever
01RetrieverThis is the output port through which the node delivers retrieved content to a root node. Drag from the retriever socket of the root node and pick Workflow Retriever, and the chain will call your workflow whenever it needs context.

This is the output port through which the node delivers retrieved content to a root node. Drag from the retriever socket of the root node and pick Workflow Retriever, and the chain will call your workflow whenever it needs context.

Key parameters

  • Required : a root node with a retriever port will not run without something connected there.
  • Question and Answer Chain : the usual pairing, where the chain asks a question and this node supplies the material to answer it.
  • MultiQuery Retriever : sits between the chain and this node, rephrasing a question several ways before calling it.
  • Contextual Compression Retriever : also accepts it, and trims what comes back before the chain sees it.
Use cases
a question and answer chain that answers from live order data, with the order lookup kept in its own workflow.
Parameters

Which parameters does the node expose?

The Workflow Retriever node has 8 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Source

source

What you see in n8n

Notes & use cases

Tells n8n which workflow to call when the chain asks for documents. Everything else on the node describes what to send that workflow, so this is the first field to set.

Key parameters

  • Source : database (Database) loads the workflow from the database by ID, parameter (Parameter) loads it from a parameter.
  • Workflow ID : required with Database, the identifier of the workflow to execute.
  • Workflow JSON : required with Parameter, the full workflow JSON code to execute, pasted into the node.
Use cases
keep a retrieval workflow in the database while you iterate on it, and reach for Workflow JSON only when you want a self-contained copy travelling with the node.
02

Name

fields.name

What you see in n8n

Notes & use cases

Names one of the values handed to the called workflow. The called workflow reads it like any incoming field, so the name you type here is the name it expects on its side.

Key parameters

  • Name : name of the field to set the value of, with dot-notation supported, as in data.person[0].name; the panel shows e.g. fieldName until you type.
Use cases
a workflow that filters customers by segment expects a field called segment, so that is what goes here, and the value follows in the matching value field.
03

Type

fields.type

What you see in n8n

Notes & use cases

Declares what kind of value the field carries. Picking a type here changes which value field appears below, which is why the node lists a separate value parameter per type.

Key parameters

  • Type : stringValue (String), numberValue (Number), booleanValue (Boolean), arrayValue (Array), objectValue (Object).
Use cases
a lookup keyed on an account identifier stays on String, while a flag that tells the workflow to include archived records belongs on Boolean.
04

Value

fields.stringValue

What you see in n8n

Notes & use cases

Holds the text a field carries once Type is set to String. Most retrieval parameters end up here: identifiers, category names, language codes, a search term copied from the question.

Key parameters

  • Value : a string, typed literally or built from an expression such as {{ $json.customerId }} when the value comes from the item that entered the root node.
Use cases
passing a product reference so the called workflow returns only the documentation pages attached to that product.
05

Value

fields.numberValue

What you see in n8n

Notes & use cases

Carries a numeric value for a field whose Type is Number. The node still accepts it as a string in the panel, so an expression that resolves to a number works the same way as a typed digit.

Key parameters

  • Value : a string field for the number to send, often an expression like {{ $json.limit }} rather than a fixed figure.
Use cases
handing the called workflow how many rows to bring back, so the same retrieval workflow serves a short answer and a long one.
06

Value

fields.booleanValue

What you see in n8n

Notes & use cases

Sends a yes or no to the called workflow when Type is Boolean. Useful for switches the workflow reads in an If node before choosing a branch.

Key parameters

  • Value : true (True) or false (False), picked from the selector rather than typed.
Use cases
a flag that tells the retrieval workflow whether to include drafts, so an internal chain sees them and a customer-facing one does not.
07

Value

fields.arrayValue

What you see in n8n

Notes & use cases

Passes several values at once under a single name when Type is Array. The called workflow receives them as a list and can loop over them or use them as a filter set.

Key parameters

  • Value : a string field for the list, written in the shape the placeholder shows, e.g. [ arrayItem1, arrayItem2, arrayItem3 ].
Use cases
sending the tags a question touches so the workflow queries every matching folder in one go instead of being called repeatedly.
08

Value

fields.objectValue

What you see in n8n

Notes & use cases

Carries a structured payload when Type is Object. This is the field to reach for when the called workflow needs a small set of related values rather than one flat entry.

Key parameters

  • Value : a json field, so the content is written as JSON and can nest keys inside it.
Use cases
forwarding a filter object holding a date range and a status, which the called workflow drops straight into its API query.
Need help

Need help automating Workflow Retriever with n8n?

A person reads every message.

FAQ

Questions builders ask about this node

01Is the Workflow Retriever node included in n8n?
Yes. It belongs to the AI package shipped with n8n, so there is nothing to install and nothing extra to pay on the n8n side. It behaves the same on n8n Cloud, the hosted offer run by n8n, and on an instance you host yourself with Docker or npm under the Community Edition and its Sustainable Use licence. A workflow built in one place runs in the other without changes. The only billing that shows up around an AI chain comes from the provider behind the chat model you connect, which charges for its own API calls under its own terms.
02What do you need to set up before the n8n Workflow Retriever node works?
Nothing on the account side. This node has no credential and no Authentication selector, so there is no key to create and no permission to grant for the node itself. What it does need is a workflow to call, reachable either by ID with Source on Database or pasted in as workflow JSON with Source on Parameter, and a root node with a retriever port to attach to. Authentication only enters the picture through the chat model sub-node connected to that root node, which carries the provider credential, and through whatever the called workflow queries.
03What are the limits of the Workflow Retriever node?
It has no main input and never runs on its own, so you cannot test it by clicking it in isolation: it executes when the root node asks for context. The node exposes 8 parameters and none of them control how many documents come back, how they are ranked or whether results are cached, because that responsibility sits inside the workflow you call. Filtering, deduplicating and trimming all belong there. This page documents version 1, the highest in the catalog, and an older workflow may show a shorter panel.
04What should you connect to it, and where do you start?
Start with a Question and Answer Chain as the root node and attach this node to its retriever port. The chain also needs a chat model sub-node, the node that talks to the model provider, and that sub-node holds the API key. Then build the called workflow separately until it returns clean items for a test input, because a retriever is only as good as what comes back from it. MultiQuery Retriever and Contextual Compression Retriever accept this node too, and both are easier to reason about once the plain chain already answers correctly.
05n8n or Make for this kind of AI retrieval?
It depends on where your data can live and how you want to pay. n8n runs on your own server or on n8n Cloud, so a retrieval workflow that touches sensitive records can stay inside your network. Make is hosted only, billed per operation, which is predictable for steady volumes but adds up when a chain calls a retriever on every question. On the building side, n8n gives you a sub-node model where a whole workflow can serve as a retrieval step, and Make favors linear scenarios. Pick on hosting, data control, cost model and how you like to build.
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