Descending networks transform command signals into population motor control. This study reveals that command-like descending neurons (comDNs) in Drosophila recruit additional descending neurons (DNs) to orchestrate complex behaviors requiring the control of multiple body parts. ComDNs, previously thought to drive behaviors alone, co-activate larger DN populations. Connectome analyses and experimental manipulations show that this recruitment is explained by direct excitatory connections between comDNs and interconnected DN networks in the brain. Descending population recruitment is essential for behavioral control: DNs with many downstream partners require network co-activation to drive complete behaviors, while those with fewer partners only elicit simple movements. These DN networks reside in behavior-specific clusters that inhibit each other. The findings support a model where behaviors are generated by recruiting increasingly large DN networks that combine multiple motor subroutines. The study also shows that comDNs connect to other DNs, forming larger DN networks. The necessity of recruiting downstream DN networks varies depending on the behavior, with some behaviors requiring network recruitment while others do not. The study further reveals that DN networks form excitatory clusters associated with distinct actions that inhibit each other. These results suggest a new framework reconciling two dominant models of DN control: comDNs drive complete behaviors by recruiting additional downstream DN populations, which combine and coordinate multiple motor subroutines. The study also highlights the importance of DN connectivity in behavior generation, with different behaviors associated with different DN clusters. The findings have implications for understanding descending control in other species, including mammals, and suggest new avenues for designing flexible artificial controllers in engineering and robotics.Descending networks transform command signals into population motor control. This study reveals that command-like descending neurons (comDNs) in Drosophila recruit additional descending neurons (DNs) to orchestrate complex behaviors requiring the control of multiple body parts. ComDNs, previously thought to drive behaviors alone, co-activate larger DN populations. Connectome analyses and experimental manipulations show that this recruitment is explained by direct excitatory connections between comDNs and interconnected DN networks in the brain. Descending population recruitment is essential for behavioral control: DNs with many downstream partners require network co-activation to drive complete behaviors, while those with fewer partners only elicit simple movements. These DN networks reside in behavior-specific clusters that inhibit each other. The findings support a model where behaviors are generated by recruiting increasingly large DN networks that combine multiple motor subroutines. The study also shows that comDNs connect to other DNs, forming larger DN networks. The necessity of recruiting downstream DN networks varies depending on the behavior, with some behaviors requiring network recruitment while others do not. The study further reveals that DN networks form excitatory clusters associated with distinct actions that inhibit each other. These results suggest a new framework reconciling two dominant models of DN control: comDNs drive complete behaviors by recruiting additional downstream DN populations, which combine and coordinate multiple motor subroutines. The study also highlights the importance of DN connectivity in behavior generation, with different behaviors associated with different DN clusters. The findings have implications for understanding descending control in other species, including mammals, and suggest new avenues for designing flexible artificial controllers in engineering and robotics.