The file type is detected automatically: single-cell, Visium spatial and bulk count matrices each get the quality metrics that belong to them — a spot is not a cell, and a doublet estimate on spatial data would be meaningless.
Cannot share the data? Use the no-upload self-check — five numbers from the Cell Ranger summary, computed in your browser.
Upload a count matrix and get back the QC every tool gives you — per-cell distributions, data-driven thresholds (median ± 3 MAD, not copied from a tutorial), a barcode-rank plot and the genes that dominate your counts — and then the analysis itself: clustering, a UMAP and automated cell-type annotation of your own data, typically within two minutes. No account, no email required. Your file is deleted once the report is built — in practice a couple of minutes after the analysis
Genes, UMIs and mitochondrial fraction per cell, with the suggested cut-offs drawn on top.
Computed from your own distributions rather than the standard 200 / 2500 / 5% that fails on many tissues.
Missing mitochondrial annotation, undetected empty droplets, suspiciously shallow profiles, over-aggressive filtering.
Not just QC: your data clustered, embedded and labelled by cell type — the same pipeline that starts a paid project, run on a subsample so it finishes while you wait.
Everything above as a document for your supervisor or core facility. No attribution required.