Any server with Grabbit installed acts as a Grabbit peer that can send, or receive content to another Grabbit peer.
To pull content into a server, a new job needs to be created on the receiving server. To do this, using the RESTful API exposed by Grabbit, PUT /grabbit/job with configuration specifying server to pull from, paths to pull, etc. This is outlined in more detail at Running Grabbit
A recommended systems layout style is to have all content from a production publisher copied down to a staging "data warehouse" (DW) server to which all lower environments (beta, continuous integration, developer workstations, etc.) will connect. This way minimal load is placed on Production, and additional DW machines can be added to scale out if needed, each of which can grab from the "main" DW.