Abstract Background Manchurian walnut (Juglans mandshurica Maxim.) is a tree with multiple industrial uses and medicinal properties in the Juglandaceae family (walnuts and hickories). J. mandshurica produces juglone, which is a toxic allelopathic agent and has potential utilization value. Furthermore, the seed of J. mandshurica is rich in various unsaturated fatty acids and has high nutritive value. Findings Here, we present a high-quality chromosome-scale reference genome assembly and annotation for J. mandshurica (n = 16) with a contig N50 of 21.4 Mb by combining PacBio high-fidelity reads with high-throughput chromosome conformation capture data. The assembled genome has an estimated sequence size of 548.7 Mb and consists of 657 contigs, 623 scaffolds, and 40,453 protein-coding genes. In total, 60.99% of the assembled genome consists of repetitive sequences. Sixteen super-scaffolds corresponding to the 16 chromosomes were assembled, with a scaffold N50 length of 33.7 Mb and a BUSCO complete gene percentage of 98.3%. J. mandshurica displays a close sequence relationship with Juglans cathayensis, with a divergence time of 13.8 million years ago. Combining the high-quality genome, transcriptome, and metabolomics data, we constructed a gene-to-metabolite network and identified 566 core and conserved differentially expressed genes, which may be involved in juglone biosynthesis. Five CYP450 genes were found that may contribute to juglone accumulation. NAC, bZip, NF-YA, and NF-YC are positively correlated with the juglone content. Some candidate regulators (e.g., FUS3, ABI3, LEC2, and WRI1 transcription factors) involved in the regulation of lipid biosynthesis were also identified. Conclusions Our genomic data provide new insights into the evolution of the walnut genome and create a new platform for accelerating molecular breeding and improving the comprehensive utilization of these economically important tree species.