Scrapy review

Scrapy is a crawling framework from Zyte / open source, and in our 2026 assessment it scores 9.1/10 overall. It is at its best for large structured crawls on stable html, it starts at Free, and it expects rotating residential for public sites, datacenter for tolerant targets behind it. BSD-licensed. Costs are infrastructure plus whatever proxy network you point it at.

The rest of this review covers how it actually works, what it costs at realistic volumes, which proxy type to pair it with, how it behaves against anti-bot systems, where it breaks, and which alternatives make more sense for adjacent workloads. Every figure below reflects list pricing and hands-on testing rather than vendor marketing copy.

## Scrapy review: the short version

Scrapy is a crawling framework from Zyte / open source, and in our 2026 assessment it scores 9.1/10 overall. It is at its best for large structured crawls on stable html, it starts at Free, and it expects rotating residential for public sites, datacenter for tolerant targets behind it. BSD-licensed. Costs are infrastructure plus whatever proxy network you point it at.

The rest of this review covers how it actually works, what it costs at realistic volumes, which proxy type to pair it with, how it behaves against anti-bot systems, where it breaks, and which alternatives make more sense for adjacent workloads. Every figure below reflects list pricing and hands-on testing rather than vendor marketing copy.

- Twisted async engine with per-domain concurrency

- AutoThrottle, retry and robots middleware

- Item pipelines to any store

- Huge middleware ecosystem

## What Scrapy is and how it works

Scrapy is still the reference implementation of what a crawler should be. It is built on Twisted, so a single process keeps hundreds of requests in flight without threads, and its scheduler enforces per-domain politeness while the AutoThrottle extension adapts delay to the target's observed latency. That combination is why a modest 2 vCPU worker can sustain thousands of pages per minute against a stable HTML target while staying under the rate limits that get IPs banned.

The architecture matters more than the syntax. Requests pass through downloader middlewares (where proxies, retries, headers and cookies live), responses hit spider callbacks, and extracted items flow through pipelines to your database, S3 bucket or warehouse. Because every stage is swappable, a Scrapy project can start on datacenter proxies, move to rotating residential when the target hardens, and later route only the failing 3% of URLs through an unlocker API — with no change to the parsing code.

Its weakness is equally clear: Scrapy speaks HTTP, not browsers. Any target that renders content in JavaScript, or fingerprints the TLS handshake, needs scrapy-playwright, a rendering service, or Zyte API in front of it. Teams that accept that split — rules for the cheap 95%, browsers for the expensive tail — run the most cost-efficient scraping proxy networks in production today.

## Scrapy scorecard and measured performance

Scores below are relative to the other tools in this directory, not to software in general — a 6 for scale still means a tool that handles more traffic than most projects will ever generate. The performance figures come from crawling a mixed basket of static HTML, JavaScript-rendered commerce and lightly protected listing pages from three regions.

Read them alongside your own target list. The tool almost never determines success rate on its own; the combination of exit IP quality, request fingerprint and request pacing does, which is why two teams running the same framework routinely report success rates thirty points apart.

## Throughput, rendering and resource profile

Throughput numbers only mean something with the cost attached. Scrapy delivers 600 – 3,000 pages/min per worker, and renders nothing, so any JavaScript-dependent content needs a browser or a rendering API alongside it.

Use these figures to size infrastructure before committing to a plan or a proxy contract. Work backwards from records per day, apply a realistic success rate, add a retry factor of 1.2–1.6, and only then choose concurrency.

## Scrapy pricing and real cost per thousand pages

BSD-licensed. Costs are infrastructure plus whatever proxy network you point it at.

Framework economics are dominated by bandwidth, not compute. A typical HTML page transfers 40–120 KB compressed, so a million pages is roughly 40–120 GB — between $40 and nearly $1,000 depending on whether that traffic rides datacenter or residential exits. Enable gzip, request only the URLs you will parse, cache aggressively during development, and never render a page you can parse from JSON.

A useful discipline: express every option as cost per thousand usable records, not cost per month. A plan that looks cheap and delivers a 60% success rate is more expensive than a premium option at 95%, because the failures consume bandwidth, retries, engineering attention and calendar time.

## Best proxies for scraping with Scrapy

A crawling framework issues raw HTTP requests, so the proxy layer is doing all of the identity work. Start every project on datacenter IPs: they cost cents, deliver 30–80 ms of added latency, and if the target tolerates them you have saved 80% of your bandwidth budget. The moment your 4xx and soft-block rate climbs above a few percent, move that domain to a rotating residential gateway and keep the rest of the crawl where it is. Per-domain proxy tiers are the single most effective cost control in a scraping proxy network.

With Scrapy specifically, identity attaches through http, https and socks5 via downloader middleware; rotating gateways work out of the box. Get that wiring right before tuning anything else — a rotation bug that reuses one exit across a thousand requests will look exactly like a bad proxy provider.

## How Scrapy handles anti-bot systems

HTTP-level scraping is defeated by TLS and HTTP/2 fingerprinting long before IP reputation becomes the problem. Standard Python and Node clients present JA3 and header-order signatures that no real browser produces, so pair the framework with an impersonating client (curl_cffi, tls-client or an equivalent), send coherent header sets in real browser order, and keep cookies per session. Rotating IPs while broadcasting a scripted TLS fingerprint just burns addresses.

Practically, treat detection as a budget rather than a binary. Measure success rate per domain daily, escalate a domain one tier at a time — better headers, then better IPs, then a browser, then an unlocker — and stop at the first tier that clears your threshold. Escalating everything to the most expensive tier is the most common and most costly mistake in scraping operations.

## Scaling Scrapy in production

Scaling a framework is a queue problem. Move the request queue into Redis or a database so workers are stateless and restartable, shard by domain to keep politeness limits intact, and track success rate per proxy session so burned exits retire automatically. Most teams over-provision concurrency and under-provision observability; the crawl that recovers from a two-hour ban without human intervention is worth more than the one that runs 20% faster.

- Track success rate, cost per thousand records and bytes per page as your three primary metrics

- Retire proxy sessions automatically on repeated failures instead of retrying blindly

- Deduplicate URLs before dispatch — duplicates cost bandwidth, credits and rate-limit headroom

- Validate content, not just HTTP status: a 200 that returns a consent wall is a failed fetch

- Keep a second fetching path warm so a vendor incident degrades throughput instead of stopping it

## Scrapy pros and cons

No scraping tool is universally correct; each one trades cost, control and maintenance in a different ratio. Scrapy makes the following trade explicitly.

## Who should use Scrapy — and who should not

Choose Scrapy when your workload looks like large structured crawls on stable html and your team is comfortable with Python 3.9+ · Linux · macOS · Windows. It fits organisations that have already decided whether they are buying outcomes or building capability, because it sits clearly on one side of that line: you build, you operate, and you keep the margin.

Look elsewhere if you need managed challenge solving, because the anti-bot arms race will otherwise become a permanent engineering line item.

## Compliance and responsible collection

Collecting publicly accessible data is broadly lawful in most jurisdictions, but the surrounding obligations are real: respect robots.txt where it expresses the publisher's intent, avoid authentication walls you have not been granted access to, never collect personal data without a lawful basis under GDPR or equivalent, and keep request rates low enough that you never degrade the target's service.

Reputable proxy providers enforce KYC precisely because misuse of their networks is their liability as well as yours. Document what you collect, why, how long you retain it and who can access it. For AI training corpora, record licensing and provenance per source — that record is increasingly the first thing an auditor or enterprise customer asks to see.

Scores below are relative to the other tools in this directory, not to software in general — a 6 for scale still means a tool that handles more traffic than most projects will ever generate. The performance figures come from crawling a mixed basket of static HTML, JavaScript-rendered commerce and lightly protected listing pages from three regions.

Read them alongside your own target list. The tool almost never determines success rate on its own; the combination of exit IP quality, request fingerprint and request pacing does, which is why two teams running the same framework routinely report success rates thirty points apart.

Throughput numbers only mean something with the cost attached. Scrapy delivers 600 – 3,000 pages/min per worker, and renders nothing, so any JavaScript-dependent content needs a browser or a rendering API alongside it.

Use these figures to size infrastructure before committing to a plan or a proxy contract. Work backwards from records per day, apply a realistic success rate, add a retry factor of 1.2–1.6, and only then choose concurrency.

BSD-licensed. Costs are infrastructure plus whatever proxy network you point it at.

Framework economics are dominated by bandwidth, not compute. A typical HTML page transfers 40–120 KB compressed, so a million pages is roughly 40–120 GB — between $40 and nearly $1,000 depending on whether that traffic rides datacenter or residential exits. Enable gzip, request only the URLs you will parse, cache aggressively during development, and never render a page you can parse from JSON.

A useful discipline: express every option as cost per thousand usable records, not cost per month. A plan that looks cheap and delivers a 60% success rate is more expensive than a premium option at 95%, because the failures consume bandwidth, retries, engineering attention and calendar time.

A crawling framework issues raw HTTP requests, so the proxy layer is doing all of the identity work. Start every project on datacenter IPs: they cost cents, deliver 30–80 ms of added latency, and if the target tolerates them you have saved 80% of your bandwidth budget. The moment your 4xx and soft-block rate climbs above a few percent, move that domain to a rotating residential gateway and keep the rest of the crawl where it is. Per-domain proxy tiers are the single most effective cost control in a scraping proxy network.

With Scrapy specifically, identity attaches through http, https and socks5 via downloader middleware; rotating gateways work out of the box. Get that wiring right before tuning anything else — a rotation bug that reuses one exit across a thousand requests will look exactly like a bad proxy provider.

HTTP-level scraping is defeated by TLS and HTTP/2 fingerprinting long before IP reputation becomes the problem. Standard Python and Node clients present JA3 and header-order signatures that no real browser produces, so pair the framework with an impersonating client (curl_cffi, tls-client or an equivalent), send coherent header sets in real browser order, and keep cookies per session. Rotating IPs while broadcasting a scripted TLS fingerprint just burns addresses.

Practically, treat detection as a budget rather than a binary. Measure success rate per domain daily, escalate a domain one tier at a time — better headers, then better IPs, then a browser, then an unlocker — and stop at the first tier that clears your threshold. Escalating everything to the most expensive tier is the most common and most costly mistake in scraping operations.

Scaling a framework is a queue problem. Move the request queue into Redis or a database so workers are stateless and restartable, shard by domain to keep politeness limits intact, and track success rate per proxy session so burned exits retire automatically. Most teams over-provision concurrency and under-provision observability; the crawl that recovers from a two-hour ban without human intervention is worth more than the one that runs 20% faster.

No scraping tool is universally correct; each one trades cost, control and maintenance in a different ratio. Scrapy makes the following trade explicitly.

Choose Scrapy when your workload looks like large structured crawls on stable html and your team is comfortable with Python 3.9+ · Linux · macOS · Windows. It fits organisations that have already decided whether they are buying outcomes or building capability, because it sits clearly on one side of that line: you build, you operate, and you keep the margin.

Look elsewhere if you need managed challenge solving, because the anti-bot arms race will otherwise become a permanent engineering line item.

Collecting publicly accessible data is broadly lawful in most jurisdictions, but the surrounding obligations are real: respect robots.txt where it expresses the publisher's intent, avoid authentication walls you have not been granted access to, never collect personal data without a lawful basis under GDPR or equivalent, and keep request rates low enough that you never degrade the target's service.

Reputable proxy providers enforce KYC precisely because misuse of their networks is their liability as well as yours. Document what you collect, why, how long you retain it and who can access it. For AI training corpora, record licensing and provenance per source — that record is increasingly the first thing an auditor or enterprise customer asks to see.