What Zyte is: a scraping API that owns proxies
Zyte, formerly Scrapinghub, maintains Scrapy, the Python crawling framework, and its commercial products sit on that heritage: Zyte API for fetching and extraction, and Scrapy Cloud for deploying and scheduling spiders in a hosted environment. Proxies exist inside those products rather than beside them.
Zyte API is billed per successful response. The cost of a request is set by an automatic difficulty tier assigned to the target site and by the features you enable, such as browser rendering, screenshots, network capture or automatic extraction. Requests that fail or are rate-limited are not charged. The IP layer is deliberately abstracted: by default Zyte chooses datacenter or residential itself, based on what the target appears to require.
That is a coherent product, and it explains why most people searching for a Zyte alternative are not looking for a cheaper Zyte. They are usually looking for something further down the stack: the IPs, without the engine wrapped around them.
Zyte Smart Proxy Manager: closed to new signups
Zyte's standalone proxy product has a long lineage. It started as Crawlera and was later renamed Smart Proxy Manager. Zyte's documentation now states that Zyte API is replacing it and that it is no longer possible to sign up to Smart Proxy Manager, with a published migration guide for existing users.
If you used that product simply as a rotating proxy behind your own crawler, this is a change of shape rather than a change of price. Zyte API can still be used through a proxy interface on api.zyte.com:8011, with an HTTPS variant on port 8014, authenticating with your API key as the proxy username and an empty password, so existing plumbing largely survives. But proxy mode drops several Zyte API capabilities by design, including server-managed sessions, automatic extraction, screenshots and network capture, and cookies can only be set for the target URL's own domain.
So the honest framing is this: if your original reason for using Zyte was "I need IPs, I already have a crawler", then the natural replacement is a proxy network, not another scraping API.
Zyte API vs raw proxies: structural differences
The differences worth acting on are structural, not matters of degree. Each row below is a different unit of purchase, not a better or worse version of the same thing.
- Billing unit: Zyte API charges per successful response, priced by the target's difficulty tier and the features enabled. Nefo charges per gigabyte transferred, at a rate fixed by the plan you bought.
- Targeting: Zyte API documents geolocation at country level, using two-letter country codes, split into standard and extended country sets. Nefo targets 227 countries and territories, US states and cities, region and city in selected countries such as the UK, Canada, Germany and France, and specific ASNs or ISPs within a country.
- Session semantics: Zyte's sessions are client-managed, using a UUID you supply, or server-managed through a session context Zyte maintains, and exist mainly to hold ban-evasion state. Nefo sticky sessions hold one IP for a duration you configure, from one minute up to 24 hours.
- Protocols: Zyte API's proxy interface is documented for HTTP and HTTPS. Nefo issues HTTP, HTTPS and SOCKS5 credentials, all with username and password authentication.
- Residential access: Zyte requires you to complete its KYC procedure before you can pin ipType explicitly to residential, and doing so changes the cost of the request. Nefo is a residential network by default.
- Mobile: Zyte API's mobile device setting is browser emulation, not a carrier IP. Nefo rents 4G/5G ports on real carrier modems by duration, each with one live rotating mobile IP you can hold or change on demand.
How Nefo proxies are bought and run
Nefo is prepaid bandwidth. You buy a volume, starting at 5GB, and generate as many proxy credentials as you like at no charge; only traffic passing through them consumes bandwidth. Flexible plans are non-expiring, so unused bandwidth stays in the account until it is used rather than resetting at the end of a month. There is no spend limit to negotiate and no monthly commitment to reach a given rate.
Targeting and session behaviour are set on the credential itself rather than per request, which is what makes the swap into an existing crawler cheap. You choose the protocol, the country and, in the United States, the state or city, or an ASN paired with a country. You then choose rotating, which gives a new IP on each request, or sticky, which holds one IP for anything from a minute to 24 hours. Usage is tracked in the dashboard against the credential.
The mobile line is separate and priced differently. Rather than paying per gigabyte, you rent a 4G/5G port on a real carrier modem for a duration from hours up to 30 days, with US state- and carrier-level targeting available. That is a different product from Zyte's, and worth mentioning only because carrier IPs are the one requirement Zyte's stack does not address at all.
How to move from Zyte to Nefo
The mechanical part is small. If you were using Zyte API in proxy mode, you are already routing through an HTTP proxy, so you are changing a host, a port and a credential. The authentication shape differs slightly: Zyte takes the API key as the proxy username with an empty password, whereas Nefo credentials are a conventional username and password pair, with the geography and session behaviour encoded into the credential you generated.
The part that takes real work is everything Zyte was doing on your behalf and no longer will.
Plan a period where both run in parallel over a slice of the same URL set, and compare success rate and cost per usable response rather than cost per request. Those are the numbers that decide it.
- Retries and backoff move to you. Zyte does not charge for failed responses; on a metered proxy network, a retried request consumes bandwidth whether or not it succeeded.
- Fingerprinting moves back into your client, whether that means a TLS-impersonating HTTP client, a hardened headless browser build, or simply disciplined header and cookie handling.
- Nothing renders JavaScript for you. If a target requires a browser, you run the browser and route it through the proxy.
- Automatic extraction has no equivalent. You keep, or rewrite, your own parsers and selectors.
- Re-express your budget in bytes. Sample the average compressed response size for your targets, multiply by expected request volume, then add headroom for redirects and retries.
- Reduce bytes deliberately: keep gzip or brotli enabled, block images, fonts and media in headless browsers, and use a plain HTTP request wherever a full browser fetch is not required.
When to stay with Zyte and when to switch
Zyte remains the better fit for a recognisable set of teams, and switching away from it for the wrong reason will cost more than it saves.
- Stay with Zyte if your bottleneck is bans on genuinely hard targets rather than IP supply. Absorbing that problem is the product, and rebuilding it in-house is not a weekend of work.
- Stay if per-success billing matters to how you forecast. On difficult targets where a meaningful share of attempts fail, not paying for failures is a real structural advantage.
- Stay if you already run Scrapy Cloud and value having deployment, scheduling and fetching under one account and one support relationship.
- Stay if you need browser actions, screenshots or structured automatic extraction, none of which a proxy network provides.
- Consider Nefo if you have your own crawler and want IPs priced predictably per gigabyte rather than variably per request.
- Consider Nefo if you need geography below country level, such as a specific US state or city, or a named ASN, which Zyte's documented country-code geolocation does not express.
- Consider Nefo if you need SOCKS5 for tools that are not HTTP clients, long sticky sessions with an explicit duration, or mobile 4G/5G ports on real carrier modems.
- Consider Nefo if you want to test with a few dollars of non-expiring bandwidth rather than configuring spend limits or agreeing a monthly commitment.
Frequently asked questions
Largely, at the transport layer. Both are HTTP proxies, so you change host, port and credentials. Zyte uses your API key as the proxy username with an empty password, while Nefo uses a normal username and password. Per-request controls like geolocation move into username tokens such as -country-us. The real work is replacing Zyte's ban handling and retries.
Yes. On Nefo you pay for bytes transferred whether or not the response was the one you wanted, so blocks, CAPTCHAs and retries consume bandwidth. If your stack already succeeds reliably, a fixed per-GB rate is usually cheaper and easier to forecast. Measure cost per usable response on a sample first. See pricing.
No. Nefo is a proxy network: it gives you IPs with protocol, geography and session control, but does not render JavaScript, run browser actions, take screenshots or parse pages. If those are load-bearing, keep Zyte for those targets or run your own headless browsers through the proxy.
Zyte documents geolocation with two-letter country codes, so targeting is at country level, and pinning ipType to residential requires completing its KYC procedure. Nefo targets down to US state and city, 22,417 cities in all, and can pin an ASN or ISP. See US city-targeted proxies.
Zyte's pay-as-you-go tier has no monthly commitment at its lowest spending limit; raising the limit involves a monthly spend. Nefo is prepaid: the smallest plan is 5 GB, bandwidth never expires, and you top up when needed. Credentials are free, so the only commitment is the bandwidth itself.
Zyte API's proxy interface is documented for HTTP and HTTPS, on api.zyte.com:8011 with an HTTPS variant on port 8014; SOCKS5 is not part of it. Nefo residential proxies support HTTP, HTTPS and SOCKS5. If you proxy something that is not an HTTP client, that is usually decisive. See SOCKS5 proxies.