Stochastic approximation Markov Chain Monte Carlo (SAMCMC) algorithms are a class of online algorithms having wide-ranging applications, particularly within Markovian systems. In this work, we study the optimization of steady-state queueing systems via the general perspective of SAMCMC. Under a practical and verifiable assumption framework motivated by queueing systems, we establish a key characteristic of SAMCMC, namely the Lipschitz continuity of the solution of Poisson equations. This property helps us derive a finite-step convergence rate and a regret bound for SAMCMC's final output. Leveraging this rate, we lay the foundation for a functional central limit theory (FCLT) pertaining to the partial-sum process of the SAMCMC trajectory. This FCLT, in turn, inspires an online inference approach designed to furnish consistent confidence intervals for the true solution and to quantify the uncertainties surrounding SAMCMC outputs. To validate our methodologies, we conduct extensive numerical experiments on the efficient application of SAMCMC and its inference techniques in the optimization of GI/GI/1 queues in steady-state conditions.