Why we exist
In late 2023, a major consumer genomics company disclosed that hackers had accessed the genetic data of nearly 7 million people. The company that had asked millions to "trust us with your DNA" could not protect it. By 2025, it had filed for bankruptcy, and the fate of 15 million people's genetic records hung in the balance.
Meanwhile, the science of genomics was accelerating faster than ever. AI systems like AlphaFold were predicting protein structures at a pace unimaginable a decade ago. A pan-European trial (PREPARE, Swen et al., The Lancet 2023) showed that a pharmacogenetic panel could reduce clinically relevant adverse drug reactions by about 30%. The technology to read and interpret human DNA was ready. What was missing was a layer that tools could call without each of them rebuilding it.
DeepDNA was founded to fill that gap: a genomic knowledge layer, built in Europe, that AI agents and other software can call, and that never asks for anyone’s genome.
Our mission
Your DNA is a letter your body wrote before you were born. Most people will read it through the tools they already use: a personal agent, a health app, a clinic's software. We give those tools cited records to read it against, without the file ever reaching us.
DeepDNA is built for the people building the tools, and today it is three things, all free with no key and no account: a remote MCP server at https://deepdna.ai/mcp, the same records as a REST API, and deepdna-mcp, an open-source package that reads a 23andMe, AncestryDNA or MyHeritage/FamilyTreeDNA file on the user’s own computer and sends nothing from the file to DeepDNA. We used to plan a consumer report; we retired it in September 2026 to become the layer that agents call instead. DeepDNA does not parse DNA files or lab reports on its servers.
European origin, European values
DeepDNA is based in Spain and operates under Spanish and EU law. This is not incidental; it is fundamental to how we operate. Europe wrote the GDPR, and we believe it should also lead the era of genomic self-knowledge rather than import it.
Being European-born means:
- GDPR-native architecture — We did not retrofit privacy. The EU General Data Protection Regulation informed every technical decision from day one. Genetic data is classified as a special category under GDPR Article 9, and we treat it accordingly.
- Data sovereignty — DeepDNA does not parse DNA files on its servers and does not ask for genetic data; a raw DNA file is read by the local package, on the user’s own computer.
- Regulatory alignment — We design for the GDPR and for the AI Act's transparency obligations, and we will say which European Health Data Space provisions apply to us when they apply. We design for compliance, not against it.
- Cultural responsibility — From the inscription at the Temple of Delphi — Know thyself — to Kant's Sapere Aude, the European tradition of self-knowledge runs deep. We carry it forward in nucleotides.
The team
DeepDNA is built by a small team in Spain, with AI assistance and human review on everything that is published, including this page. We are intentionally lean, and we would rather describe what the work involves than inflate who does it.
Curation
Reading the primary literature and the guideline bodies, and turning them into records where every claim has a source and every empty field stays empty rather than guessed.
The local package
The unglamorous part: provider formats, genome builds, strand orientation and no-calls, handled on the user’s own computer. A missing or unreadable position is reported as unknown, never as normal. This is where the correctness lives.
Privacy and infrastructure
Keeping the servers free of personal data: no genotype accepted by the MCP server, no DNA file on DeepDNA’s side, no cookies, and lookups counted only as short labels.
Writing
Docs, gene pages and the blog: explaining what a result means and, just as carefully, what it does not.
Everyone shares a single conviction: your genome belongs to you.
How we work
How the API works
The knowledge endpoints serve a curated dataset — 16 genes, 39 variants and 83 drug–gene pairs today, each record with its sources — kept in a versioned database. Every day an automated watcher checks CPIC, ClinVar, FDA drug labels, AlphaMissense and new papers for changes that affect them; PharmGKB, PharmVar, DPWG, OMIM and dbSNP are not checked automatically yet. A person reviews every change before it is published, and each one is listed in the changelog. No third-party call and no model in the response path. The sources are ClinVar, PharmGKB, PharmVar, the CPIC and DPWG guidelines, OMIM, dbSNP and the primary literature.
The MCP server serves the same records to AI agents as tools, and takes no genotype. The local package deepdna-mcp downloads the same public records for every user and compares them with a raw DNA file on the user’s own computer. DeepDNA does not parse DNA files or lab reports on its servers. For the approach to variant interpretation and evidence, see our Methodology page.
Privacy by design
The knowledge endpoints and the MCP server receive nothing personal, and no DNA file reaches DeepDNA: the local package reads it on the user’s computer and sends nothing from it to us. We set zero cookies and use no third-party trackers. Our Privacy Policy describes what the website collects.
Our values
- Privacy by design — Your genetic data is the most personal information that exists. We engineered our entire system around not keeping it.
- Scientific rigour — Every claim in a DeepDNA record is backed by a cited source. We say when a field is empty rather than guessing. We never overstate what the science says.
- Transparency — We tell you exactly what we do with data, how the API works, and where every record comes from. No black boxes. Read our Methodology for the full technical approach.
- Spanish and EU law — DeepDNA operates from Spain, under Spanish and EU law, and designs for the GDPR rather than around it.
- Accessibility — Genomic knowledge should not be a luxury. Everything is free: the knowledge endpoints, the MCP server and the local package need no key and no account. No paywall on what a gene means.
- Open science — We publish our methodology and cite every data source. Science that cannot be scrutinised is not science.
What we are not
DeepDNA is not a medical device and does not provide medical diagnoses. The API’s responses are educational information. It does not replace a clinician; it gives the tools you use something concrete to bring to one.
We do not collect DNA samples. We do not run laboratory tests. We do not sell reports to individuals. We do not parse DNA files or lab reports on our servers. We publish cited records that the agents people already use can look up, and a local package that reads the genotype file people already own on their own computer. Think of it as a reference that other software calls: the DNA file stays with its owner, and the record says what a position means, with the source a clinician can check.
Transparency pages
We believe that trust is built through transparency. These pages document every aspect of how DeepDNA operates:
- Methodology — How the records are written from guidelines and the literature, how they are checked, and what they do not do
- Privacy Policy — What data we collect, how we handle genomic data, GDPR compliance, and your rights
- Terms of Service — The legal terms governing use of DeepDNA
- Manifesto — What we believe and why we exist
- API Docs — Every endpoint, the MCP server and the local package, with the response shape and data handling
Contact
For general enquiries, partnerships, or press:
- Email: hello@deepdna.ai
For privacy-related questions or data rights requests:
- Email: privacy@deepdna.ai
For scientific or methodology questions:
- Email: science@deepdna.ai