Fluid inclusions trapped in coesite-bearing rocks provide important information on the fluid phases present during ultrahigh-pressure metamorphism. The subduction-related coesite-bearing eclogites of the Tso Morari Complex, Himalaya, contain five major types of fluids identified by microthermometry and Raman spectroscopy. These are: (1) high-salinity brine, (2) N2, (3) CH4, (4) CO2 and (5) low-salinity aqueous fluids. These fluids were trapped during both deep subduction and exhumation processes. The coesite-bearing rocks are inferred to have been buried to a depth of >120 km, where they experienced ultrahigh-pressure metamorphism. The fluid–rock interaction provides direct evidence for fluid derivation during a deep subduction process as demonstrated by silica–carbonate assemblages in eclogite. High salinity brine, N2 and CH4 inclusions are remnants of prograde and peak metamorphic fluids, whereas CO2 and low-salinity aqueous fluids appear to have been trapped late, during uplift. The high-salinity brine was possibly derived from subducted ancient metasedimentary rocks, whereas the N2 and CH4 fluids were likely generated through chemical breakdown of NH3-bearing K minerals and graphite. Alternatively, CH4 might have been formed by a mixed fluid that was released from calcareous sediments during subduction or supplied through subducted oceanic metabasic rocks. High density CO2 is associated with matrix minerals formed during granulite-facies overprinting of the ultrahigh-pressure eclogite. During retrogression to amphibolite-facies conditions, low-salinity fluids were introduced from external sources, probably the enclosing gneisses. This source enhances salinity differences as compared to primary saline inclusions. The subducting Indian lithosphere produced brines prior to achieving maximal depths of >120 km, where fluids were instead dominated by gaseous phases. Subsequently, the Indian lithosphere released CO2-rich fluids during fast exhumation and was then infiltrated by the low-salinity aqueous fluids near the surface through external sources. Elemental modelling may improve quantitative understanding of the complexity of fluids and their reactions.