Extension, Desktop App or Cloud: How to Pick a No-Code Scraper

There are 24 no-code scrapers in this directory, and most comparisons of them rank the tools against each other on features. That is the wrong first cut. Point-and-click selection, auto-detected tables, pagination handling and CSV export are close to table stakes now; nearly every product in the category does all four adequately.

What differs is where the scraper runs. That one choice decides whether the job can be scheduled, whether it survives you closing your laptop, whether it can reach a page behind your login, what happens when the target starts blocking you, and whether the output can reach another system without a person carrying a file. Sort the category by deployment model first and the shortlist usually collapses to two or three candidates before you have compared a single feature.

Three places a scraper can run

In your browser, as an extension. The scraper is a Chrome or Edge add-on that reads the DOM of the tab in front of you. It uses your session, your cookies and your home or office IP address. It runs while you watch it and stops when you close the tab.

On your machine, as a desktop application. The scraper is a program with its own embedded browser. It has a task list, a scheduler tied to your operating system, and it writes files to local disk. It runs on one computer's IP address and only while that computer is awake.

On someone else's machine, as a cloud platform. You build the job in a web app or a companion extension, then the vendor executes it on their infrastructure. Their proxies, their concurrency, their clock. You get an API and a bill per run.

Everything else follows from those three sentences.

Browser extensions

Extensions are the fastest path from "I need this table" to a CSV on your desktop, and they are the only model that solves the logged-in-page problem cleanly.

Instant Data Scraper is the purest version. It is a Chrome and Edge extension with no account and no plan. You open a page, click the icon, and it guesses which repeating block on the page is the data. Its Chrome Web Store listing, read on 27 July 2026, states that "this extension is completely FREE" and describes automatic navigation to the next page via buttons or links, support for infinite scrolling, delay and maximum wait time customisation, and data export to an Excel spreadsheet or CSV file. There is no configuration to save and no project to name. It is free because the work happens in your browser, so nobody is paying for cloud compute.

Data Miner adds structure. It works from "recipes", which are saved extraction rules, and it ships a public library of community recipes for common sites so you often do not build one at all. Its pricing page, read on 27 July 2026, offers a free tier of 500 pages per month and carries an unusually blunt warning: exceed 500 pages in a month and the account locks indefinitely until you upgrade to a paid plan. That is worth knowing before you start a job you cannot pause. Data Miner also states that scraped data never touches its servers, which matters if you are pulling from a system your employer would rather not send to a vendor.

Listly is the spreadsheet-shaped one. It runs in Chrome, Edge and Whale, you click page elements to define what to pull, and it pushes straight into Google Sheets rather than making you download and re-import a file. Its free tier is a demonstration rather than a working allowance. A pricing review by Prospeo, which says it verified the numbers against Listly's own pricing page in early 2026, puts the free plan at 10 URLs a month with a 5-row export cap and the entry-level Light plan at $30 a month, or $15.60 a month if paid annually.

Web Scraper straddles two models and is worth understanding as two products. The extension is free, runs inside Chrome's DevTools panel, and works from a "sitemap": a saved plan describing how to walk a site and what to take at each level. That sitemap is portable, which is what makes the extension unusual. The same definition runs unchanged on Web Scraper Cloud, so the extension doubles as a free authoring environment for a paid cloud job.

Extensions are wrong for you if anything about the job repeats on a schedule. There is no scheduler, because there is no process to schedule. They are also wrong for high-volume work against a site you care about your account on. Every request carries your session and your IP, so an anti-bot system that decides you are scraping penalises the account you logged in with, not an anonymous pool address. The flip side is that your ordinary residential IP and real browser fingerprint are exactly what gets an extension through pages that a datacenter request bounces off, which is why extensions still work on targets that defeat a naive script.

Desktop applications

A desktop app is what you buy when the job repeats, you want to own the tool outright, and the data should stay on your disk.

WebHarvy is the clearest expression of the model. It is a Windows point-and-click scraper that renders pages like a browser, handles logins, form submission, pagination and keyword lists, and is sold as a perpetual licence rather than a subscription. Its purchase page on 27 July 2026 lists a single-user licence at $99, discounted from a $129 list price, with multi-user and site licences above that. Each licence includes one year of updates and email support plus permanent access to versions released in that window. The vendor's current marketing also advertises connecting to local or hosted language models, through Ollama, LM Studio or the commercial APIs, to pull fields out of unstructured text.

ScrapeStorm is the cross-platform option, running on Windows, macOS and Linux, and it offers two authoring modes: a Smart Mode that infers the data pattern automatically and a Flowchart Mode where you draw the navigation steps. Its free entry point is genuinely unclear from the outside. The vendor's own FAQ describes a 14-day free trial with no credit card, while a Thunderbit review describes a permanently free Starter tier limited to 10 tasks, one concurrent run and 100 exported rows per day, with a Professional personal plan at $45 a month. Those two accounts do not reconcile. Confirm which one applies before you plan around a free tier.

Octoparse is the hybrid that trips people up. It is a desktop client, historically Windows-first, with an auto-detect workflow builder, but its scheduling and concurrency live in a paid cloud tier. On the free plan you are running a desktop scraper: local execution, no cloud runs, no scheduling. Octoparse's pricing page, read on 27 July 2026, shows a free forever plan alongside Standard at $69 and Professional at $249 with a 16 percent annual discount applied; the company's June 2026 pricing explainer gives the undiscounted monthly figures as $83 and $299, which reconciles. The free plan is capped at 10 tasks on one device, local extraction only, two concurrent local runs and 50,000 exported rows a month. Third-party trackers quote Standard anywhere between $75 and $99, which tells you how much to trust any figure in a listicle. Take the vendor's number and check it yourself on the day you buy.

Desktop apps are wrong for you if the job needs to run while the machine is asleep, if more than one person needs to run or edit it, or if you are scraping at a volume where one IP address is the bottleneck. A desktop scheduler is only as reliable as the laptop it lives on, and a laptop that closes at 6pm does not run a 2am job. They are also wrong for anyone who cannot install software, which in a lot of companies is most people.

Cloud platforms

Cloud is where you go when the scraper has to become infrastructure: running on a clock, at concurrency, behind proxies, feeding another system.

Browse AI is built around monitoring rather than one-off extraction. You train a "robot" by walking through a site once in a companion extension, then the platform re-runs it on a schedule and tells you what changed. Pricing is credit-based. Its pricing page, read on 27 July 2026, lists four tiers: a free plan with 50 credits a month, Personal at $48 a month or $19 a month billed annually, Professional at $87 a month or $69 a month billed annually, and a Premium tier from $500 a month billed annually. Fifty credits is sized for validating a workflow, not running one. Credit pricing rewards annual prepay heavily enough that the monthly and annual numbers look like different products.

Apify is the platform with the highest ceiling and the steepest learning curve in this group. Its unit is the Actor, a packaged scraping program, and its store holds thousands of them, so for a well-known target the job is often picking an existing Actor rather than building anything. Billing is usage-based on compute units, roughly one gigabyte-hour of runtime each, layered on a monthly plan that includes prepaid credit. A PriceTimeline capture dated 13 January 2026 records a free tier carrying about $5 of monthly platform credit, Starter at $29 a month, Scale at $199 and Business at $999; documentation from Use Apify verified against the vendor's pricing page on 18 May 2026 shows the same shape. Other secondary sources put Starter at $39 or $49, so treat the tier structure as reliable and any single number as needing a check. Apify is also the point where "no-code" becomes a spectrum: the store is no-code, writing your own Actor is not.

Web Scraper Cloud is the narrower, more predictable option and the easiest upgrade path if you already built a sitemap in the free extension. Its pricing page on 27 July 2026 lists Project at $50 a month with 5,000 URL credits, two concurrent scrapers and 30-day retention; Professional at $100 with 20,000 credits and three concurrent scrapers; Scale from $200 with unlimited URL credits and 60-day retention; and a custom Enterprise tier. Scheduling and API access are included on every paid tier, which is unusual here and makes the $50 plan the cheapest cloud option in this guide that a data team can pull from programmatically. The extension stays free for local work.

Cloud platforms are wrong for you if the target is behind a login you cannot safely hand to a vendor, if the job runs once and never again, or if nobody on the team can own a usage bill. Credit-based billing punishes exploratory work, because a misconfigured job that retries a thousand pages costs the same as a correct one. They also add the one failure mode extensions never have: the platform's IP addresses are shared and known, so a target that tolerates your browser may block their pool.

The models side by side

Fit by job shape

The decision usually resolves faster if you describe the job rather than the tool.

The hand-off question people skip

Ask what the tool gives you at the end, and who has to touch it.

A file means a person is in the loop forever. Extensions almost all end here, and for a one-off that is correct. It becomes a problem the fourth time somebody emails a CSV to the analytics team.

A spreadsheet push is the middle ground and it is underrated. Listly's Google Sheets integration removes the download and re-import step, and a sheet is something a warehouse or a BI tool can read on a schedule without anyone opening a browser. For a small team this is often enough automation.

An API or webhook is what a data team wants. Apify stores every run's output in a dataset you retrieve programmatically, and Web Scraper Cloud ships API access on all four of its paid tiers. A webhook flips the direction: instead of polling for new results, the platform tells your system that a run finished. If the eventual destination is a database, decide up front whether the scraper pushes or your pipeline pulls, because that single decision rules out most of the category.

There is also a durability question that does not show up in feature lists. A job configured in an extension lives in one browser profile on one laptop. When that person changes machines or leaves, the job is gone unless someone exported the config. Cloud jobs live in an account you can add people to. If the scrape is going to matter in six months, that is not a small difference.

Where the money goes

Four pricing shapes exist in this category, and they fail in different ways.

Free and local, as with Instant Data Scraper, costs nothing because you are supplying the compute. The hidden cost is your attention: someone sits and watches it.

Capped free tiers are the most common shape and the most likely to surprise you. Data Miner's 500 pages a month with an account lock on overrun is the honest version, because the limit is stated plainly. Read the cap and the overrun behaviour before you start, not after.

One-time licences are rare enough now to count as a differentiator. WebHarvy's perpetual licence with a year of updates means a job you run for three years costs what it cost on day one. The trade is that support and updates lapse, and the tool cannot grow past one machine.

Credits are how every cloud platform bills, whether the unit is a URL, a robot run or a compute unit. Credits are honest for variable workloads and hostile to experimentation, because a broken selector that retries costs the same as a working one. If you go this route, run your first month deliberately small and measure the credit burn on one real job before you commit annually.

Every number in this guide is attributed and dated because prices in this category move constantly, and the third-party trackers disagree with the vendors more often than they agree. Octoparse's Standard tier was quoted at four different prices across the sources consulted for this guide. Check the vendor's own page on the day you buy.

When no-code is the wrong category entirely

Three signals say the job has outgrown this category regardless of deployment model.

The first is blocking. If your targets are actively fighting you, switching to a different point-and-click tool will not fix it. What fixes it is a proxy network underneath the tool, or a scraping API that handles rotation and rendering for you. Our guide to choosing a web scraping tool covers that decision, and the Bright Data vs Oxylabs comparison covers the enterprise end of it.

The second is volume with variation. Once you are scraping hundreds of sites whose layouts differ, maintaining hundreds of point-and-click configurations costs more than writing one parser. That is the point where a crawling framework starts paying for itself, and the 2026 stack overview sketches how the layers fit together.

The third is anything with a latency requirement. No-code tools are batch tools. If a product feature depends on fresh data within seconds, you are building a service, not running a scraper.

A short decision path

Answer four questions in order and stop at the first one that decides it.

1. Does it need to run when you are not there? If yes, cloud. Nothing else schedules reliably.
2. Is the data behind a login you cannot share with a vendor? If yes, extension or desktop. The credentials stay on your side.
3. Does the output have to reach another system automatically? If yes, cloud with an API or webhook, and check the tier that includes it.
4. Is this a one-off? If yes, the free extension. Do not buy anything.

If none of the four decides it, the job is probably a weekly refresh on a handful of sites, which is exactly the overlap where all three models work. In that case pick on billing preference: a perpetual licence if you hate subscriptions, an extension plus a calendar reminder if you want free, cloud credits if your time is worth more than the bill.