Your DNA, decoded on your own machine.

gnomancer turns the raw file you already downloaded from AncestryDNA, 23andMe, MyHeritage, FamilyTreeDNA or LivingDNA into one readable report: lineages, traits, carrier status, drug metabolism, and a plain-language summary. Open source. Runs locally. Nothing is uploaded.

Install in 5 minutes How to get your raw data Source on GitHub

# macOS / Linux (Windows via WSL): one line to install, one to run
$ curl -fsSL https://gnomancer.com/install.sh | bash
$ gnomancer ~/Downloads/AncestryDNA.zip

A gnomancer report: overview tiles for vendor, sites, sex, Y haplogroup, mtDNA haplogroup and blood group, with quality checks

The report, generated from a synthetic demo file. Real reports open with an at-a-glance digest and an optional narrative written by your own AI assistant.

What you get

One self-contained HTML report (plus Markdown, TSV and JSON), built from the tools researchers actually use.

Lineages

Y haplogroup from yhaplo with a second opinion from Yleaf on current YFull, FTDNA and ISOGG trees; mitochondrial haplogroup from Haplogrep3. Each comes with population context.

Traits and ancestry markers

A curated panel of about 110 well-replicated sites: eye and hair pigmentation, lactase persistence, caffeine and alcohol metabolism, bitter taste, blood group, and more.

Health and carrier sites

APOE, Factor V Leiden, hemochromatosis, celiac HLA tags, macular degeneration, type 2 diabetes and the rare pathogenic sites that happen to be on consumer chips. Collapsed by default, for people who would rather not look.

ClinVar scan

Every typed site checked against ClinVar. Each hit is scored for plausibility, because arrays are unreliable at rare sites and a homozygous rare "pathogenic" call is usually an artifact.

Pharmacogenomics

PharmCAT star-allele diplotypes and CPIC phenotypes for 18 genes, with the drugs each one matters for and an honest per-gene coverage figure.

Plain language, and your own AI

An automatic at-a-glance digest, plus a narrative written by whichever AI assistant you already use. gnomancer is an MCP server too: connect it to Claude Desktop, Claude Code or Cursor and just ask "what does my report say about caffeine?" The file still never leaves your machine.

How it works

Three steps, all on your computer.

Download your raw data

Every major vendor lets you export it from your account settings. Step-by-step for each company. gnomancer reads the zip directly and detects the vendor and genome build.

Run one command

The installer sets up the open-source tools and public reference files once. Then gnomancer yourfile.zip does everything else in about a minute.

Read the report

Open REPORT.html. Light and dark, printable, searchable, with the technical detail one scroll below the summary.

Reads AncestryDNA23andMeMyHeritageFamilyTreeDNALivingDNAGenes for GoodtellmeGenVCF and remaps older builds to GRCh37.

Privacy, stated plainly

Your genotype file never leaves your computer. There is no account, no upload, no telemetry.

The only network traffic is the one-time download of public tools and reference data (yhaplo, Haplogrep3, PharmCAT, ClinVar, the rCRS mitochondrial sequence).
The optional --ai summary sends the structured result tables, never the raw file, to the Anthropic API, and only when you ask for it.
This website is static. It sets no cookies and runs no analytics.

Under the hood

gnomancer is glue around tools that have been validated by the people who built them. Every one is open source.

Tool	Contributes	Source
snps	vendor detection, normalization, build remapping	apriha/snps
yhaplo	Y haplogroup	23andMe/yhaplo
Yleaf	Y haplogroup on current trees	genid/Yleaf
Haplogrep3 + PhyloTree	mitochondrial haplogroup	genepi/haplogrep3
ClinVar	clinically classified variants	NCBI
PharmCAT	CPIC / PharmGKB star-allele calling	PharmGKB/PharmCAT
ezancestry	1000 Genomes population estimate	arvkevi/ezancestry

Not medical advice. A genotyping array reads a fixed set of common sites, about 0.02% of the genome. It cannot see rare variants, copy-number genes or full HLA types, and its calls at rare sites are wrong often enough that a published check found 40% of them did not confirm (Tandy-Connor et al., Genetics in Medicine 2018). gnomancer tells you what the chip says and how much to trust it. Anything that matters deserves a clinical-grade test and a conversation with a clinician or genetic counselor.