Since it was published, Coxhead's (2000) Academic Word List (AWL) has been frequently used in English for academic purposes (EAP) classrooms, included in numerous teaching materials, and re-examined in light of various domain-specific corpora. Although well-received, the AWL has been criticized for ignoring some important facts that words still tend to show irregular distributions and are used in different ways across disciplines (Hyland & Tse, 2007). One such difference concerns collocations. Academic words (e.g. "analyze" and "concept") often co-occur with different words across domains and sometimes even refer to different meanings. What EAP students need, accordingly, is a "discipline-based lexical repertoire" (Hyland & Tse, p.235). Inspired by Hyland & Tse's insightful remarks, we developed an online corpus-based tool, "TechCollo," which is meant for EAP students to explore collocational knowledge in a domain or compare collocations across disciplines. TechCollo runs on textual data stored in three specialized corpora and utilizes frequency and some information-theoretical measures (e.g. mutual information) to decide whether co-occurring word pairs constitute collocations. In this article we describe the current version of TechCollo and how to use it in EAP studies. Particularly, we report a pilot study in which we employed TechCollo to investigate whether the AWL words take different collocates in different domain-specific corpora. This pilot basically confirmed Hyland & Tse's indications and demonstrated that many AWL words show uneven distributions and collocational differences across disciplines.