Mobile　edge　computing　(MEC)　provides　a　fresh　opportunity　to　significantly　reduce　the　latency　and　battery　energy　consumption　of　mobile　applications.　It　does　so　by　enabling　the　offloading　of　parts　of　the　applications　on　mobile　edges,　which　are　located　in　close　proximity　to　the　mobile　devices.　Owing　to　the　geographical　distribution　of　mobile　edges　and　the　mobility　of　mobile　devices,　the　runtime　environment　of　MEC　is　highly　complex　and　dynamic.　As　a　result,　it　is　challenging　for　application　developers　to　support　computation　offloading　in　MEC　compared　with　the　traditional　approach　in　mobile　cloud　computing,　where　applications　use　only　the　cloud　for　offloading.　On　the　one　hand,　developers　have　to　make　the　offloading　adaptive　to　the　changing　environment,　where　the　offloading　should　dynamically　occur　among　available　computation　nodes.　On　the　other　hand,　developers　have　to　effectively　determine　the　offloading　scheme　each　time　the　environment　changes.　To　address　these　challenges,　this　paper　proposes　an　adaptive　framework　that　supports　mobile　applications　with　offloading　capabilities　in　MEC.　First,　based　on　our　previous　study　(DPartner),　a　new　design　pattern　is　proposed　to　enable　an　application　to　be　dynamically　offloaded　among　mobile　devices,　mobile　edges,　and　the　cloud.　Second,　an　estimation　model　is　designed　to　automatically　determine　the　offloading　scheme.　In　this　model,　different　parts　of　the　application　may　be　executed　on　different　computation　nodes.　Finally,　an　adaptive　offloading　framework　is　implemented　to　support　the　design　pattern　and　the　estimation　model.　We　evaluate　our　framework　on　two　real-world　applications.　The　results　demonstrate　that　our　approach　can　aid　in　reducing　the　response　time　by　8%–50%　and　energy　consumption　by　9%–51%　for　computation-intensive　applications.　©　2019,　Science　China　Press　and　Springer-Verlag　GmbH　Germany,　part　of　Springer　Nature.