The Zyte alternative, now that Smart Proxy Manager is gone

Smart Proxy Manager was Zyte's proxy product, and the company announced it was being retired for new customers. The current notice is short: the product is retired and no longer available, with Zyte API named as its replacement, and its documentation page opens with a warning that it is no longer open to new sign-ups, with existing users pointed to a migration guide.

The replacement, Zyte API, is not a proxy service with a new name. It is a scraping API: you send it a URL and it decides which proxies to use and whether a browser is needed, and it can parse the result too. Proxies are still in there, and the pricing page lists residential, datacenter and mobile IPs with automatic rotation, included in the API. But you do not buy them, and by default you do not choose them.

For a working Smart Proxy Manager integration, migration is real work rather than a credential swap. You need a separate Zyte API key, because a Smart Proxy Manager key does not carry over, and the closest thing to the old endpoint is a proxy mode at api.zyte.com:8011, authenticated with a Zyte API key.

A proxy seller could wave this away. A good part of it is true, so we will not.

Zyte's blog published an article arguing that the proxy era is ending as scraping shifts from managing IP pools to API-driven tools. The case is that sites now look at far more than the address: the TLS handshake, the browser fingerprint, how the page is driven. A clean exit with a fingerprint that gives you away still fails. That is correct, and it is why a proxy alone is not a scraping strategy.

The cost argument rests on its own analysis that more than 80% of Zyte API users spend the same as or less than they would on proxy-based setups. That is a vendor's analysis of its own customers, and nobody outside can audit it. The point underneath still holds: if your team spends its week fixing bans instead of using data, paying someone to absorb that work can cost less than the bandwidth it replaces.

Where we part company is the conclusion. A scraping API is a proxy with the decisions made for you. That suits teams who want them made, not teams who already make them well or run their own browsers.

These two models do not line up neatly, and neither is simply cheaper.

Zyte API charges per successful request, and the rate depends on how hard the site is. On pay as you go it is $0.13 to $1.27 per 1,000 requests for a plain HTTP response, across five website tiers, and $1.01 to $16.08 per 1,000 requests for a browser-rendered page. Zyte also offers monthly commitments of $100, $200 or $500 that lower the per-request rate, with enterprise volume pricing through sales; at the top tier the rate starts from $0.06 per 1,000 successful responses. A failed request costs nothing, which is a real advantage on a hostile target.

We charge per gigabyte, whatever the outcome: $1.50/GB on Essential, $3.00/GB on Premium and $3.00/GB on Mobile. A retry costs traffic. In return we do not tier the site you visit, and a light page costs very little.

Compare them on your own pages: take a week of traffic, count requests and gigabytes, and price both. Light pages from easy sites tend to favour per-gigabyte billing. Heavy pages from hard sites that need rendering can favour success billing, and if they do, buy from Zyte. Our sheet is on the pricing page.

This is the real difference between the products, and it matters more than price.

Zyte API makes an automatic switch between device residential and datacenter proxies based on what each site requires. Its own residential page describes residential proxies as a method of last resort, used only if absolutely needed. If you want a residential exit on a site where Zyte would pick datacenter, its documentation says forcing a residential IP on a request requires business verification and affects the request cost. Location works the same way: you get a country of origin, picked automatically or set per request, with the extended country set affecting cost, drawn from residential IPs in more than 200 countries.

With us, the exit type is the product. Buy Essential or Premium and every request leaves from a residential address; buy Mobile and it leaves from a phone on a carrier network. Country, US state and ASN go in the username on every pool, and Premium and Mobile add city.

Neither approach is wrong. With varied targets and trust in the algorithm, automatic choice saves money. If you are testing how a site treats one carrier in one city, you need to be the one choosing.

Zyte's documentation is candid about one thing worth knowing before you build on it: a session moves to a new IP in the same country when its residential address expires. That suits a scraping API, where the country matters more than the address, but a session there is not a promise of one IP.

Our rule is published the same way on every page: up to 24 hours (120 minutes on Essential). Premium and Mobile hold the longest, and within the window the exit does not move, which is what a multi-step flow behind a sign-in needs.

The client side differs too. Zyte offers a proxy mode at api.zyte.com:8011, authenticated with a Zyte API key, and its own page lists the gaps: no screenshots, browser actions, automatic extraction or server-managed sessions in proxy mode. The documented setup also calls for installing Zyte's CA certificate for some HTTP clients, Python's requests among them, and the migration guide adds that proxy mode is not optimized for use with browser automation tools. Ours is an ordinary proxy: HTTP on port 7777, SOCKS5 on port 1080 on Premium, Mobile and ISP, and HTTPS through a standard CONNECT tunnel with no certificate to install.

Zyte's catalogue is Zyte API, the managed Zyte Data service, Scrapy Cloud hosting, an MCP server and the open-source Scrapy framework. That is a strong stack for teams that scrape, and Scrapy in particular is a gift to the field.

Ours goes toward infrastructure instead. Static ISP subnets are $255 a month for a /24, each owned by one customer, and Mac servers are from $32 a day or $160 a month. Zyte sells neither, and we sell no crawling, rendering or parsing. See ISP proxies and Mac servers.

If you were a Smart Proxy Manager customer and what you really wanted was data, follow Zyte to Zyte API. It does more than the old product did, it is billed on success, and the trial costs nothing. We mean that.

If what you wanted was a proxy, a line you paste into any client with an exit you chose, Zyte no longer sells that and says it does not believe in it. That is our product: residential, mobile, static subnets and Macs, with the session rule published on every page.

So the real question is whether you want someone else's algorithm deciding how your requests reach the web. If yes, Zyte. If no, start with a small wallet here and test it on your own targets.

Zyte appears across review sites and benchmarks, now mostly as a scraping API. We quote none of their scores; a copied figure is a claim we cannot keep current.

Our own record is on the trust page, every Zyte fact on this page links to the page we read it on, and our prices are read live rather than typed in.