To test direct and indirect effects of glomalin, mycorrhizal hyphae, and roots on aggregate stability, perspex pots separated by 37-μm nylon mesh in the middle were used to form root-free hyphae and root/hyphae chambers, where trifoliate orange (Poncirus trifoliata) seedlings were colonized by Funneliformis mosseae or Paraglomus occultum in the root/hyphae chamber. Both fungal species induced significantly higher plant growth, root total length, easily-extractable glomalin-related soil protein (EE-GRSP) and total GRSP (T-GRSP), and mean weight diameter (an aggregate stability indicator). The Pearson correlation showed that root colonization or soil hyphal length significantly positively correlated with EE-GRSP, difficultly-extractable GRSP (DE-GRSP), T-GRSP, and water-stable aggregates in 2.00–4.00, 0.50–1.00, and 0.25–0.50 mm size fractions. The path analysis indicated that in the root/hyphae chamber, aggregate stability derived from a direct effect of root colonization, EE-GRSP or DE-GRSP. Meanwhile, the direct effect was stronger by EE-GRSP or DE-GRSP than by mycorrhizal colonization. In the root-free hyphae chamber, mycorrhizal-mediated aggregate stability was due to total effect but not direct effect of soil hyphal length, EE-GRSP and T-GRSP. Our results suggest that GRSP among these tested factors may be the primary contributor to aggregate stability in the citrus rhizosphere.