Some embodiments provide a capacity management agent that modifies bandwidth that is allocated between an end user and a carrier network by caching requested content that is streamed at a first rate and then providing the cached content to the end user through the carrier network at a second rate. The agent performs a process that includes receiving data intended for a service region of the carrier network from an external data network. The process identifies resource availability at the service region. Next, the process passes the data to the service region at the first rate when the resource availability at the service region is not less than a threshold amount and caches the data for passing to the service region at the second rate that consumes fewer carrier network resource than the first rate when the resource availability at the service region is less than the threshold amount.