Residential proxies, billed by the gigabyte

A residential proxy exits through an address an internet service provider assigned to a consumer connection. The target sees a subscriber on a home line — the same kind of address its own customers arrive on — instead of a range registered to a hosting company.

That is the whole advantage, and it is a real one. Reputation on the public internet is assigned by range. Hosting ranges are published, catalogued and scored. Consumer ranges are treated more carefully, because blocking one costs the operator paying customers.

90M+ residential addresses sit behind a single hostname. You do not buy a list, refresh it, or replace entries as they die. You send a request to gw.anonedge.com and the gateway resolves the exit. Read list versus gateway if you are migrating off a static list.

You buy traffic and you spend traffic. What you do not spend stays on the balance, because credits carry over indefinitely. That is where one residential proxy provider differs from another, rather than from the subscription vendors: an allowance that resets on the thirtieth is a use-it-or-lose-it charge wearing the costume of a rate. Non-expiring traffic means a quiet month costs nothing.

Two things about the meter, stated here rather than in a footnote. It counts every byte on the wire — request headers, TLS negotiation, redirects and the responses you did not want, not only the page body you kept. And traffic routed through Target Filters, meaning state, city, ZIP or ASN selection, bills at twice the standard rate. Country selection and ASN exclusion are included at the base rate.

Neither is visible on a rate card, and both change the arithmetic. What providers bill you for breaks the meter down line by line, and reducing bandwidth cost covers what to strip before it moves.

Every plan carries a 7-day money-back guarantee. Buy the smallest amount of traffic that answers your question and run it against the target you actually care about.

Port 823 carries HTTP and HTTPS. Port 824 carries SOCKS5. Both are rotating endpoints, where the gateway resolves the exit for you. Ports 10000 through 20000 are sticky: one address held for a window you choose between 1 and 120 minutes, 30 minutes by default.

Pick by task, not by preference. Anything that reads a page and moves on runs rotating. Anything that holds a login, a cart or a multi-step form runs sticky, because a session that changes address halfway through is a session the target has every reason to drop. The mechanics are in connecting to the gateway, and the task-by-task version is in rotating vs sticky sessions.

Both modes read from the same pool and bill at the same rate. There is no premium for holding an address, and no penalty for giving it up.

Country selection and ASN exclusion are included at the base rate. State, city, ZIP and ASN selection are available as Target Filters and bill at twice the rate — priced that way because a narrow filter draws from a narrow slice of the pool.

When no address matches the filter you set, the gateway answers 400 NO_RAY rather than quietly handing you an exit somewhere else. That is the behavior you want. A silent fallback does not fail your job; it corrupts your dataset, and you find out weeks later.

Full syntax is in targeting. Before a dataset depends on it, confirm the exit is where the filter said it was — verifying city-level targeting shows how, and ASN targeting covers selecting by network rather than by place.