<p>The manufacturing error of steel truss members features significant randomness, which may result in discrepancies between the assembly accuracy of structural members on-site and the design specifications. This paper proposes an assembly accuracy analysis approach based on error propagation and stochastic simulation, constructing a theoretical framework for analyzing and quantifying the assembly accuracy under the influence of all uncertainty factors encountered during the assembly process. In addition, the reliability of assembly accuracy, i.e., the probability that the structural assembly accuracy on-site satisfies the designed specifications, is taken as a reasonable indicator for determining whether assembly on-site is feasible. First, develop error propagation models for a steel truss and clarify errors’ propagation and accumulation mechanisms. Then, identify uncertainty parameters and sample stochastically. Finally, create assembly limit-state functions and calculate assembly accuracy and corresponding reliability through the Monte Carlo simulation. Taking a flat and a spatial steel truss in practical engineering as an example, the results show that the assembly accuracy of the flat truss is 1.361 and 1.698&#xa0;mm, and that of the spatial truss is 6.422, 6.927, 6.357, 7.346&#xa0;mm, all above accuracy with 100% reliability. The sensitivity analysis results show that the rod length error plays an important role during flat truss assembly, and so does angular error during spatial truss assembly. The developed strategy is independent of the materials and structures applied. Therefore, it can be used for more complicated structures.</p>