Aims. We perform clustering measurements of 800 X-ray selected Chandra COSMOS Legacy (CCL) Type 2 active galactic nuclei (AGN) with known spectroscopic redshift to probe the halo mass dependence on AGN host galaxy properties, such as galaxy stellar mass M_star, star formation rate (SFR), and specific black hole accretion rate (BHAR; λ_BHAR) in the redshift range z = [0−3]. Methods. We split the sample of AGN with known spectroscopic redshits according to M_star, SFR and λ_BHAR, while matching the distributions in terms of the other parameters, including redshift. We measured the projected two-point correlation function w_p(r_p) and modeled the clustering signal, for the different subsamples, with the two-halo term to derive the large-scale bias b and corresponding typical mass of the hosting halo. Results. We find no significant dependence of the large-scale bias and typical halo mass on galaxy stellar mass and specific BHAR for CCL Type 2 AGN at mean z ∼ 1, while a negative dependence on SFR is observed, i.e. lower SFR AGN reside in richer environment. Mock catalogs of AGN, matched to have the same X-ray luminosity, stellar mass, λ_BHAR, and SFR of CCL Type 2 AGN, almost reproduce the observed M_star − M_h, λ_BHAR − M_h and SFR–M_h relations, when assuming a fraction of satellite AGN f_AGN^sat ∼ 0.15 . This corresponds to a ratio of the probabilities of satellite to central AGN of being active Q ∼ 2. Mock matched normal galaxies follow a slightly steeper M_star − M_h relation, in which low mass mock galaxies reside in less massive halos than mock AGN of similar mass. Moreover, matched mock normal galaxies are less biased than mock AGN with similar specific BHAR and SFR, at least for Q > 1.