Self-serve · Updated 11 October 2026
The Ancestry Workbench: qpAdm, rotation and Fst on your own DNA
Upload the raw DNA file you already have, and the same engine that builds every Ancestrify report is yours to drive: single qpAdm models with your sample as the target, rotation batches that fit every combination of a candidate pool, and an Fst scan that ranks the whole reference panel by distance to you. You pay from a prepaid balance, only for work that completes.
Rotation row 1 · your own fit
- rank
- 1 · passing first, then p
- Anatolia_N
- 0.49 · SE 0.033
- WHG
- 0.14 · SE 0.029
- Yamnaya_Samara
- 0.37 · SE 0.036
- p
- 0.31 · not rejected
What the Workbench is
Every Ancient Ancestry report Ancestrify publishes is built the same way: your genotypes are merged into the Allen Ancient DNA Resource v66 with Poseidon's trident, qpadm() from ADMIXTOOLS 2 is run in R on our own servers, our analyst tests models, and a person chooses and approves the one that is published. The Ancestry Workbench is that same merge and that same engine with you at the controls.
You bring a kit (a raw-data export, or the sample from an Ancient Ancestry order you already have), the Workbench prepares it once, and from then on every tool below runs on your own merged genome. Nothing is smoothed or ranked for presentation: a model qpAdm rejects comes back rejected, with the p-value that says so.
It is not a subscription and it is not a report. There is no monthly plan, no credit that expires, and no analyst between you and the output. What you run is your own hypothesis, and the numbers you get back are the tool's own.
Three tools, one merged genome
- One model at a time
qpAdm, direct, on your own sample
Choose the sources and the right set from the merged panel, with your sample fixed as the target, and the model runs as qpadm() from ADMIXTOOLS 2 on the genotypes directly, never on a screening shortcut. You get the p-value and chi-square, every source's weight with its standard error, Z-score and 95% interval, and the tool's own warnings in plain words.
The same genotype-direct fit that every published Ancestrify model passes through.
costabout €0.004 per run - Every combination, fitted
Rotation, as a batch of real fits
Pick a pool of candidate sources, choose two, three or four sources per model, pin any population that must appear in every model, set the right set, and the Workbench fits every combination as its own genotype-direct qpAdm run. Rows are ranked passing first (p above 0.05 and every weight above zero), then by p-value. There is no screening step and no confirm step, because every row is already the real thing.
The results are your own hypotheses, not our published finding. A rotation is only as good as the pool you give it: its error rate depends mostly on how plausible the candidates were before testing, and the p-value alone ranks models poorly, so read the rows as a map of what the data tolerate, not as an answer.
costabout €0.004 per model, €0.04 minimum - Available before the merge finishes
Fst distance, closest populations first
Hudson's Fst between your genotypes and every population in the reference panel, ranked closest first. The closest 500 are listed, each with the number of samples behind its figure. Pseudohaploid ancient samples are handled as such, so a single-allele call is never counted twice.
It needs no merge: the scan aligns your file against the panel's markers on upload, so it is usable seconds after the kit is received, while the qpAdm tools wait for the merged bundle.
costabout €0.004 per scan
Use it from your own assistant
Connect Claude, ChatGPT, Cursor or VS Code to the Workbench, and your own assistant finds populations, runs your models, rotations and Fst scans, and reads you the results. It pays from your balance at the same prices, within a daily limit you choose, and adding a model to your report stays with you, in the dashboard.
As far as we know, Ancestrify is the first ancestry service where your own assistant runs real analyses on your DNA.
Pricing
A balance is a one-time payment. It never expires and never renews. Completed runs draw their price; failed runs are never charged.
A balance never expires, so there is no month in which it goes unused.
- Top up
- €10.00, €25.00, €50.00, €100.00, or any amount from €9.99 to €500. Paid once by card. A balance is a one-time payment. It never expires and never renews. Completed runs draw their price; failed runs are never charged.
- Kit preparation
- about €0.49, held when you add the kit and drawn only when it is Ready. A kit attached from a completed Ancient Ancestry order is free: it is already merged.
- qpAdm run
- about €0.004 per genotype-direct model, whether qpAdm accepts or rejects it. Failed runs are never charged.
- Rotation
- about €0.004 per fitted model, €0.04 minimum per batch, up to 1,600 models in one rotation. Models that fail are refunded to the balance.
- Fst distance
- about €0.004 per scan of the whole panel.
- Into your report
- €5.00 to add any finished model of your own to your Ancient Ancestry report as a new version, whatever its fit; a model with a share at or below zero cannot join. It wears the Customer model seal, and your report keeps opening on Ancestrify's version.
- Included
- The merged EIGENSTRAT bundle of every Ready kit, to download and run in your own toolchain, and a 23andMe-format export of your converted file. Neither draws from the balance.
Per-run figures are stated as about: the balance is billed to a thousandth of a euro by the server, and the exact quote for a rotation is shown before you run it. The full price list for every Ancestrify analysis is on the pricing page.
Kits and formats
A kit is one raw-data export from 23andMe, AncestryDNA, FamilyTreeDNA, MyHeritage or LivingDNA (.txt, .csv, .zip or .gz), validated by its content rather than by the vendor's name. The file is converted, filtered of indels and strand-ambiguous markers, and merged into AADR v66. The Workbench keeps the merged bundle for you; the original export can be deleted at any time.
If you already have a completed Ancient Ancestry report, its sample attaches to the Workbench as a kit at no cost: the merge was done for the report and the bundle is archived, so nothing has to be prepared again.
Whole-genome sequencing VCFs are not accepted as Workbench kits yet; the conversion currently lives inside the Ancient Ancestry order flow. You can hold up to 50 kits, and each kit fits one model at a time.
Not sure what your file supports? The free raw DNA file check reads it and says. The privacy notice names the controller and the hosting, and the source populations are listed under Ancestry.
Questions
Is this a subscription?
No. You top up a balance once, by card, and it stays yours until you spend it. It never expires, nothing renews, and there is no monthly plan to cancel. Every completed run draws its price from the balance; failed runs are never charged.
Is it real qpAdm?
Yes. Every model, including every row of a rotation, is qpadm() from ADMIXTOOLS 2 run on our own infrastructure over your genotypes merged into the Allen Ancient DNA Resource v66, with your sample as the target. There is no screening approximation and no confirm step: the fit you see is the fit that ran.
What if my model fails?
Then you learn something. A rejected model (p below 0.05, or a weight the data cannot support) is the method working: it is a completed run, qpAdm says so plainly, and it draws its price of about €0.004 like any other. A run that fails to complete, or that you cancel, is never charged, and models that fail inside a rotation are refunded to your balance.
How is a rotation different from the published analysis?
The published model on an Ancient Ancestry report is drawn from candidate sources picked for the customer's declared origins before any model runs, and a person chooses and approves it. That matters because a search's error rate depends mostly on how plausible its candidates were before testing, and the p-value alone ranks models poorly. The Workbench runs a rotation over whatever pool you choose, as a set of real fits, and labels the results as what they are: your own hypotheses, not our published finding. Read a passing row as one the data tolerate, not as the answer.
Is my data secure?
Your file is kept in private EU storage and processed on servers in Germany and Finland under GDPR, used only for your own kit, and never matched against another customer's file. You can delete a kit from the Workbench whenever no work is running on it, then the file itself from your genetic files, and deleting your account removes every kit and file with it.
How long does a kit take?
It depends on the file and on the queue. The kit card shows each stage as it happens, Received, Converting, Merging and Ready, and you are emailed when the kit is Ready. You do not have to wait for the Fst distance scan: it runs on the uploaded file directly, before the merge finishes.
Which files can I upload?
A raw-data export from 23andMe, AncestryDNA, FamilyTreeDNA, MyHeritage or LivingDNA, as .txt, .csv, .zip or .gz. Whole-genome sequencing VCFs are not accepted as Workbench kits yet. The sample from any completed Ancient Ancestry order attaches as a kit for free.
Can I use the Workbench from Claude or ChatGPT?
Yes. Connect Claude, ChatGPT, Cursor or VS Code over MCP and sign in with your Ancestrify account. Your assistant runs work from your balance at the same prices, within a daily limit you choose, and it can never add a model to your report, top up or delete a kit. The assistant page shows how to connect each one.
What happened to the Model Lab?
It became the Workbench. The Model Lab was a one-time unlock on a published Ancient Ancestry report; the Workbench is a standalone service that needs no report, adds rotation and Fst distance, and bills per run from a prepaid balance. If you had unlocked the Model Lab, your merged sample is already attached as a kit and your balance carries a credit for the unlock.
Learn the method first
The AdmixTools 2 Lab runs real f-statistics, qpWave and qpAdm over the public reference panel in the browser, free, so you can learn the workflow before preparing a kit. Run your own qpAdm models is the walkthrough, and choosing sources and outgroups is the habit that saves the most runs.
Your genome, the real engine, your hypotheses.
Top up once, prepare a kit, and run qpAdm, rotation and Fst distance for as long as the balance lasts. Failed runs are never charged.