Draft genome assemblies are commonly represented by assembly graphs. Tools for accurately aligning nucleotide and amino-acid sequences onto such graphs can facilitate hybrid assembly, read error correction, analysis of hypervariable genes, reconstruction of complex multi-domain genes and haplotype separation. At the same time, currently available solution, such as vg, GraphAligner, TAG, have significant limitations. Here, we present GAligner (GA) — a general purpose tool for local alignment of DNA sequences onto assembly graphs. In particular, GAligner is able to accurately map long erroneous sequences (e.g. single-molecule sequencing reads). We comprehensively benchmark GA using different datasets and sequencing technologies and show that it produces accurate alignments on assembly graphs of varying complexity.