WorkflowsNode ReferenceWeb & HTTP

Web Search

View as Markdown

The Web Search node runs a search on the public web and returns a list of matching pages, each with its address, title, description, and content. It finds pages; it does not interpret them. Pair it with an Agent or Web Extract when you need to turn those pages into an answer.

When to use it

  • You need current, real-world information a workflow cannot get from your own systems, like a published rate, a filing, or a recent news item.
  • You want to locate the right pages before pulling specific details from them.
  • You are verifying a business or person against public sources as part of an intake or underwriting flow.

Inputs

FieldWhat it’s forExample
Search QueryThe words to search for, just as you would type them into a search engine.”Florida Secretary of State business filing Chen Insurance Group LLC”
Max ResultsHow many pages to return, from one to ten.3
Content FormatHow much of each page to bring back: readable text, a short summary, just the links, or the raw page. Readable text is the default and works well for handing to an agent.Summary

Outputs

FieldWhat you get back
ResultsA list of matching pages. Each one includes its address, title, a short description, and the page content in the format you chose.

Example

Before quoting a new commercial account, an agency wants to confirm the business exists and find its state filing page. A Web Search node searches for the company name alongside “Secretary of State business filing” and returns the top three results. A downstream Web Extract step then opens the most relevant one and reads the filing status, so an underwriter never has to look it up by hand.

The Web Search node configuration panel, showing the search query, max results, and content format.

Web Search only finds pages, it does not read them for you. Feed its results into an Agent to pull out an answer, or into Web Extract to grab specific fields. Asking for a handful of results usually beats asking for one, since the best page is not always first.

Web Extract

Open the pages you found and pull specific fields off them.

Agent

Hand the results to an agent to summarize or extract an answer.

HTTP Request

Call a known service directly instead of searching for it.

Check

Confirm the search actually returned results before continuing.