FlowRunner
PricingContact
Theme
Start Free

FetchFox

Developer Tools

Scrape the web with FetchFox by describing the fields you want in plain English. FetchFox crawls, reads, and returns structured records, so agents collect data from sites that publish no API.

6 actions API key available
FetchFox website ↗ Platform Documentation ↗ Capability data verified 2026-08-01
A founder asks for a pricing dataset covering a competitor's catalogue
Agent calls Crawl Site with a URL pattern and explicit visit, extract and depth caps
Agent reads back the list of matching page addresses the crawl discovered
Agent counts the pages found and compares the total against the cap that was set
Agent confirms the addresses all sit under the section that was actually asked for
Agent posts the page count and a sample of the discovered addresses to the operator
Operator confirms the scope, or narrows the pattern, before Extract Data spends against it

What This Integration Enables

FetchFox's argument is not that a model can read a page. Plenty of services claim that. FetchFox's argument is that it will find the pages for you. Crawl Site takes either a URL pattern such as a products path with a wildcard, or a plain English description of the pages you want, and returns the addresses that match. Scrape Site collapses discovery and extraction into one call. That makes the interesting question on this connector a scoping question rather than a parsing question, because the thing you are describing in English is not a field, it is which pages count.

Cost follows the same logic. Billing scales with pages loaded, which is why the connector exposes Max Pages To Visit, Max Pages To Extract and Max Depth as first class controls rather than tuning knobs. Proxy tier moves the number even more sharply: FetchFox's own guidance puts residential tiers at roughly 160 times the per gigabyte cost of the free tier, and recommends starting on Automatic and escalating only when blocked. Run Agent gives all of that away deliberately, letting FetchFox decide by itself whether to crawl, extract or simply load pages and which proxy tier to use, in exchange for not having to specify anything. FlowRunner treats that trade as a decision worth surfacing, not a default worth hiding.

Without FlowRunner

Scraper as a project Getting a dataset means writing and maintaining code before you learn whether the data is even useful
Scope discovered by invoice The size of a crawl becomes clear after the bill arrives, not before it starts
Unclear page matching A pattern either over collects or misses a section, and nobody notices until the analysis is wrong

With FlowRunner

Dataset as a request Field names described in plain English return structured rows without a scraper to maintain
Bounded before it runs Visit, extract and depth caps set the ceiling on pages loaded before the first request goes out
Scope confirmed by a person The discovered URL list is reviewed before extraction spends anything against it

Use Case Scenarios

Competitor pricing dataset, scoped before it is bought

An operator wants every product page under a competitor's catalogue turned into rows. The agent runs Crawl Site first, with a URL pattern and explicit visit and depth caps, and gets back the list of matching addresses without extracting anything. It posts the count and a sample of twenty addresses for review. Once the scope is confirmed, the agent calls Extract Data across the list with a template naming each field in plain English, then writes the rows into Airtable with the source page address that Extract Data attaches to every item.

Watching one section of a site for new listings

A weekly flow calls Scrape Site against a job board or listings section with Many Items Per Page selected, because a single listing page yields several records. The agent diffs the returned items against last week's set, keeps only the new ones, and posts them to Slack. Because caps are set on the call rather than left open, a site that suddenly grows a paginated archive cannot quietly turn a weekly check into a full site crawl.

Turning a known URL list into a clean table

Someone hands over a list of two hundred company pages collected by hand. There is nothing to discover, so the agent skips crawling entirely and calls Extract Data directly with the exact URLs and a field template. For long lists it enables Run In Background, receives a job ID, and polls Get Job until the state is no longer active, treating any state other than active or completed as a failure. Finished rows land in Google Sheets, each carrying the source page address and a link to the captured HTML so a claim can be traced back to what the page actually said.

Human-in-Loop Highlight

The gate on this connector sits between discovery and extraction, because that is where an English sentence quietly becomes a bill. A semantic query like "product detail pages for laptops" is a judgment, and neither the agent nor FetchFox can tell you whether it caught the accessories section too. So the agent runs Crawl Site with caps in place, collects the addresses, and stops. It posts to the operator: "Crawl found 1,840 matching pages under example.com/products. Extracting all of them at the current proxy tier is the next call. Here are 20 of the addresses. Confirm the scope, narrow the pattern, or lower the cap?" If the crawl came back blocked and the only way forward is escalating past the Automatic proxy tier, that escalation is its own question rather than an automatic retry. FetchFox's guidance puts residential traffic at roughly 160 times the free tier's per gigabyte cost. This is FlowRunner's digital andon cord applied to spend: the agent has everything it needs to proceed, and the reason it stops is that proceeding is the part a person should own.

Agent processes routinely
Detects exception requiring judgment
Clear match Continues automatically
Ambiguous Routes to human via preferred channel
Human decides
Agent resumes with decision

Agent Capabilities

6 actions

Scraping

2
  • Scrape Site Crawls a site and extracts structured data from every matching page in one call, combining discovery and extraction. Takes a URL pattern plus a template describing the fields you want. Visit, extract and depth caps bound the cost. Runs synchronously by default and returns items inline; Run In Background returns a job ID to poll instead.
  • Crawl Site Discovers URLs without extracting anything, returning the list of matching page addresses. Targets pages either by URL pattern or by describing them in plain English as a semantic query, which requires start URLs. Used to build a reviewable URL list before any extraction spend, or to audit which pages exist under a section.

Page Content

2
  • Extract Data Pulls structured data from a known list of URLs using a template that names each field in plain English. No crawling happens, so the exact pages are passed in. Many Items Per Page suits listing pages where one page yields several records. Every item carries its source page address and a link to the captured HTML.
  • Fetch Page Content Loads a single page and returns it in whichever representations you request, with no AI extraction. Markdown, cleaned HTML, raw HTML, plain text, outbound links, images and embedded JSON-LD are available, defaulting to Markdown, HTML and Links. Used when you want the page itself rather than structured fields.

Agent

1
  • Run Agent Runs a scraping task described entirely in plain English, letting FetchFox decide whether to crawl, extract or simply load pages, and which proxy tier to use. The quickest route to a result when you do not want to specify patterns and templates, at the cost of control over scope and spend. The result may contain discovered URLs, extracted items, or both.

Jobs

1
  • Get Job Retrieves the state and, once finished, the results of a job started with Run In Background. The state field reports active while running and completed or failed once done. The agent polls until the state is no longer active and treats any state other than active or completed as a failure.

Frequently Asked Questions

What can FlowRunner do with FetchFox?

FlowRunner agents can run Scrape Site, Crawl Site, and Extract Data in FetchFox, plus 3 more actions.

Does connecting FetchFox to FlowRunner require OAuth?

No. FetchFox connects to FlowRunner with an API key, no OAuth flow required.

Can FetchFox trigger a FlowRunner workflow automatically?

FetchFox doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with FetchFox

$100 in credits. No card required. Connect in minutes.