The deepest-level relationships amongst Euglenozoa remain poorly resolved, despite a rich history of morphological examination and numerous molecular phylogenetic studies of small subunit ribosomal RNA (SSU rRNA) data. We address this question using two nuclear-encoded proteins, the cytosolic isoforms of heat shock protein 90 (hsp90) and heat shock protein 70 (hsp70). For both proteins we examined sequences from the three primary groups within Euglenozoa (euglenids, diplonemids, and kinetoplastids), and from their close relatives, Heterolobosea. Maximum likelihood (ML) and ML distance analyses of these proteins support a close relationship between diplonemids and kinetoplastids to the exclusion of the euglenid Euglena gracilis. In hsp90 and combined protein analyses bootstrap support is very strong and alternative topologies are generally rejected by 'approximately unbiased' (AU) tests. This result is consistent with recent molecular biological and morphological data, but contradicts early structural accounts and many SSU rRNA analyses that favour a closer relationship between diplonemids and euglenids. However, a re-examination of an important SSU rRNA data set highlights the instability of the inferences from this marker. The protein analyses also suggest that bodonids are paraphyletic, with trypanosomatids grouping with 'clade 2' and 'clade 3' bodonids to the exclusion of 'clade 1' bodonids.