The cuckoo search is a new bio-inspired algorithm with few parameters to be adjusted. In this paper, the parameters of the basic cuckoo search, nest number, discovered probability and step length are discussed. However, when the search scope is large the efficiency of the basic cuckoo search may decrease, so a novel improved cuckoo search is introduced to cope with the issue. Meanwhile, eight benchmark functions are employed to show the efficiency of the proposed cuckoo search. Furthermore, the proposed algorithm is utilized in designing PID controllers. Compared with the traditional Ziegler–Nichols PID (ZNPID), the controllers provided by the proposed algorithm are more effective, in that the overshoot and the transient time of the controllers are both smaller than ZNPID. While compared with particle swarm optimization and the basic cuckoo search, the convergence rate and accuracy of the proposed algorithm is better.