Resources · n8n integration

n8n SearXNG nodeConfigure SearXNG in n8n.

The n8n SearXNG node hands an AI agent a metasearch engine you host yourself. It is a sub-node, meaning it has no main input and never runs on its own: it clips onto an agent through a tool port. Four optional parameters shape what comes back, and version 1 is the one shipped today.

Try the SearXNG node on n8nSee the 4 parameters

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the n8n SearXNG node actually do?

SearXNG aggregates results from several search engines without tracking the person searching. The node exposes that instance to an AI agent as a callable tool: the agent writes the query itself, the node runs it against your instance and returns the results as JSON so the agent can read them. Nothing is configured per query, only the four options that frame every search.

Three jobs come up again and again. A research agent that answers questions about this week rather than about its training data: the model gets a search tool, decides on its own when to use it, and the Number of Results option decides how much material it has to chew on. A monitoring agent that reads the news on a schedule, summarizes it and drops the digest somewhere people already look, such as Slack or a sheet in Google Sheets. And a language-scoped watch, where Language is pinned to one two-letter code so a Spanish market review never drifts into English sources.

The node is a sub-node, a small node that plugs into a bigger one instead of sitting in the main flow. It attaches to the tool port of a root node: AI Agent, AI Agent Tool or MCP Server Trigger. Alone on a canvas it does nothing at all, and n8n will not let it execute.

When is another route better? If the workflow always runs the same query and no model is involved, an agent is overhead: you are paying a model to rewrite a string you already know. The point of this node is the decision, not the search. Give it to an agent that has to work out whether it needs to look something up, and which words to look up.

Two limits worth knowing before you build. The node needs results in JSON, a format the default SearXNG configuration does not serve, so an untouched instance answers nothing useful until its settings are changed. And the instance has to be reachable from n8n on the same network, which rules out a laptop install if n8n runs elsewhere. Beyond that, result quality tracks the health and the engine list of the instance you point at, not the node. Teams that want a walkthrough of the whole agent pattern can start with the n8n review.

Connect

How do you connect SearXNG to n8n?

  1. 01

    Get an instance running

    The node talks to a SearXNG service you run, not to a public endpoint someone else maintains. That service has to sit on a network your n8n instance can reach, so a container next to n8n is the usual setup. Where the files land depends on how SearXNG was installed, and its administrator documentation covers the install paths.

  2. 02

    Switch the instance to JSON

    SearXNG ships without JSON output, and the node needs it. Open the instance's settings.yml, find the search.formats section and add json next to html. If the section is missing entirely, create it. Restart the service afterwards, then check that a search returns JSON before going anywhere near n8n.

  3. 03

    Create the credential in n8n

    In n8n, open Credentials and pick the SearXNG credential. It holds one field, API URL, the address of the instance you want to query. A credential is stored once and reused by every workflow that needs it, so the address lives in a single place rather than in each node.

Connections

Where does the node plug in?

n8n splits AI workflows between root nodes, which sit in the flow and receive items, and sub-nodes, which hang off a root node's ports and only run when that root node calls them. SearXNG belongs to the second group, and it has exactly one port.

Output (what it plugs into)

  • Toolai_tool
01ToolThis is the output port that turns the node into something a model can call. The root node advertises the tool to the model, the model decides when a search is worth running, and the results come back into its reasoning.

This is the output port that turns the node into something a model can call. The root node advertises the tool to the model, the model decides when a search is worth running, and the results come back into its reasoning.

Key parameters

  • Required: the node has no main input and never executes on its own, so it stays inert until it is wired to a root node's tool port.
  • AI Agent: the usual host, an agent that picks its own tools while answering.
  • AI Agent Tool: an agent exposed as a tool to another agent, for nested setups.
  • MCP Server Trigger: publishes the search to clients outside n8n.
Use cases
a support agent that checks public documentation before answering, with the search tool attached next to a chat model.
Parameters

What can you tune on the SearXNG node?

The SearXNG node has 4 parameters. For each one: the node as you configure it in n8n, what the parameter changes, and our field notes.

01

Number of Results

options.numResults

What you see in n8n

Notes & use cases

Caps how many results come back from a single search. The default is 10, and everything the node returns ends up in the model's context, so this number is really a budget: more results mean more material and a longer prompt.

Key parameters

  • Number of Results: a number, left empty it stays at 10.
Use cases
a fact-checking agent that only needs a handful of corroborating links runs leaner with a low value, while a market scan that feeds a written summary benefits from a wider haul.
02

Search Page Number

options.pageNumber

What you see in n8n

Notes & use cases

Picks which page of the result list the node fetches. It starts at 1, the first page, and moving it forward skips what came before instead of adding to it. There is no paging loop built into the node, so a second page means a second call.

Key parameters

  • Search Page Number: a number, default 1.
Use cases
when the first page of a broad query is dominated by the same few sites, pointing a second SearXNG tool at page 2 gives the agent something the top results did not surface.
03

Language

options.language

What you see in n8n

Notes & use cases

Filters results by language using a two-letter code, the ISO 639 style shorthand: en for English, es for Spanish, fr for French. The default is en, which is easy to forget when the agent itself answers in another language.

Key parameters

  • Language: a string holding the two-letter code, default en. SearXNG's search syntax page lists what an instance accepts.
Use cases
a competitor watch for a Spanish-speaking market, where results in English would pull the summary off topic.
04

Safe Search

options.safesearch

What you see in n8n

Notes & use cases

Asks the engines that support it to filter explicit content. Engines that do not support the flag ignore it, so this is a preference passed along rather than a guarantee enforced by the node.

Key parameters

  • Safe Search: three choices, 0 (None), 1 (Moderate) and 2 (Strict). The default is None.
Use cases
an agent whose answers reach a public chat window is safer on Strict, since nobody reviews a search result before the model quotes it.
Need help

Need help automating SearXNG with n8n?

A person reads every message.

FAQ

Questions people ask next

01Is the n8n SearXNG 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, on n8n Cloud as well as on a self-hosted instance under the Community Edition and its Sustainable Use license. The cost that does exist sits elsewhere: the SearXNG instance runs on infrastructure you provide, and the model provider attached to the agent bills its own API calls under its own terms. n8n adds nothing to that bill.
02What do you need to connect SearXNG to n8n?
One credential with a single field, API URL, the address of the SearXNG instance you want to query. There is no authentication selector on the node and no key to rotate, because the node reaches a service you control. Two conditions matter more than the credential itself: the instance has to be reachable from n8n on the same network, and it has to serve JSON, which means adding json to the search.formats section of its settings.yml. A default SearXNG configuration does not do that.
03What are the limits of the SearXNG node?
It searches, and that is the whole scope: four options, no crawling, no page fetching, no reading of the pages it finds. Results come back as a list the model reads, and pulling the full text of a link is a separate job for another node. Version 1 is the one in the catalog, so an older workflow may show fewer options. Result quality is not the node's doing either: it depends on the engines your instance queries and on how healthy that instance is on the day.
04What has to be plugged in for the search tool to work?
At minimum a root node and a chat model. SearXNG attaches to the tool port of an AI Agent, an AI Agent Tool or an MCP Server Trigger, and that root node needs its own chat model sub-node with a provider credential, otherwise nothing reasons and nothing decides to search. Start with one agent, one chat model and this single tool. Memory and an output parser are separate ports you can add later, once you can see what the agent does with a search result.
05n8n or Make for this kind of search?
It depends on where the data is allowed to live. SearXNG exists because queries stay on an instance you run, and that argument only holds end to end if the automation platform runs next to it. n8n self-hosts with Docker or npm, so the workflow and the search engine can share a private network. Make is hosted, with no self-hosting option, and it bills per operation, which changes the arithmetic on an agent that fires several searches per answer. Both draw workflows visually, so pick on hosting, data control and cost model.
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