Longtail-SWE is a benchmark of issue-resolution tasks designed to evaluate coding agents across 13 underrepresented programming languages.