Backconnect Proxy

A backconnect proxy is a single stable gateway hostname and port that brokers each request out through a different peer in the provider's pool. You point your client at one endpoint for the life of the job, and the provider decides which exit address serves each request. Decodo's documentation, read on 2026-07-27, gives gate.decodo.com:7000 as its standard residential gateway, which is the shape every large vendor now ships.

This is not what most teams picture when they hear the word proxy. The mental model is usually a list of individual endpoints, one IP and port per line, that you load into a rotator you wrote yourself. That model still exists for datacenter and ISP products, where you are buying specific addresses. On a backconnect gateway you are buying access to a pool you never enumerate.

The consequences are operational. Rotation policy lives on the vendor side, so how often the exit changes, how peers are selected and what happens when one drops are the vendor's decisions, not yours. Geo targeting and session control move into the username string, since there is nowhere else to put them when the connection details never change. And you lose direct visibility into which address served a request. Some providers return the exit IP in a response header, but the header name is vendor specific and not universal, so confirm it in the docs before you build monitoring that depends on it. Without it, a rising block rate is hard to attribute: you cannot tell whether one bad subnet is poisoning the sample or the target has started flagging the whole pool.

The decision this informs is how much of the rotation logic you are willing to treat as a black box. For general crawling at moderate scale, accepting the vendor's policy is the right trade, since building an equivalent pool manager is real engineering for little gain. For anything where per request attribution matters, whether that is a compliance trail, a chargeback argument with the vendor, or debugging a target that blocks intermittently, ask for exit IP reporting and per request control in writing before signing. Vendors that cannot supply it are asking you to debug their infrastructure without instruments.