MyGenome
Personal Genome Exploration
README
MyGenome
Your DNA, explained in plain English. You ask the questions; the AI handles the science.
What You Can Discover
Here are some questions to start with:
"Am I a carrier for any autosomal recessive conditions?"
"Do I have variants that are associated with increased disease risk?"
"What are the variants in my database classified as 'Pathogenic' or 'Likely Pathogenic' in ClinVar that the American College of Medical Genetics recommends reporting for health reasons?"
"What does AlphaMissense predict about my missense variants — are any flagged as potentially damaging?"
"Are there any noteworthy variants in my BRCA1 or BRCA2 genes?"
"Beyond BRCA1 and BRCA2, are there any noteworthy variants in other high-penetrance cancer susceptibility genes like PALB2 or TP53?"
"Are there any variants in my genome known to affect how I metabolize common medications, specifically looking at the CYP gene family?"
"How rare or common are my variants compared to the general population?"
"Give me a breakdown of all my variants by functional consequence."
"How closely related am I to the other person in my database?"
A few example conversations show what's possible — single sample and Trio with Opus 4.6, and the same Trio with Sonnet 4.6. Log in to Claude.ai to see interactive charts.
Every question could be a starting point for a longer conversation (skipped for brevity).
This Is Not Medical Advice
Genomics is powerful but it has real limits — and so does this tool.
Most variants in your genome are of unknown significance. Even well-studied variants associated with disease risk describe probabilities, not certainties. The same variant can mean very different things in different people. And AI, however capable, can make mistakes when interpreting complex genomic information.
This tool is for personal exploration and curiosity. It is not a medical device. It does not diagnose conditions, assess your personal health risk, or replace the judgment of a doctor, genetic counsellor, or any other qualified professional.
If something in your results concerns you, please talk to a healthcare professional — ideally one with experience in genetics or genomic medicine. They can put your specific variants in the context of your personal and family history in a way no AI currently can.
Your Genome, Your Privacy
Please read this carefully. Your genome is among the most personal information that exists. It can reveal things about your health, your ancestry, and your family — not just about you, but about your relatives too. Here's what you need to know:
Your genome stays on your computer. The database where your DNA data is stored runs locally on your own machine. It does not connect to the internet. It does not "phone home" with telemetry. Nobody can access your data but you.
When you ask questions, your variants travel to the AI. When your AI assistant looks up your genome to answer a question, it sends the relevant variant data to AI service you're using. If you use a cloud-based AI like Claude or Gemini, that data is sent over the internet to the provider's servers. The AI vendor — and potentially other parties involved in processing your request — will see your genetic information.
This is a real risk. Genomic data is uniquely sensitive. Unlike a password, you cannot change your DNA if it's exposed. Think carefully about what you're comfortable sharing.
You can keep everything completely private by using a locally running AI model (Ollama). When both your genome database and your AI run on your own hardware, nothing leaves your computer.
This tool is read-only. It can look up your data to answer your questions, but it cannot change, delete, or write anything to your genome database.
How It Works
This tool connects your personal genome database to your AI assistant. When you ask a question about your DNA, the AI looks up your actual variant data and gives you a real answer based on your genome — not a generic textbook response. It can search by gene, by clinical significance, by how rare a variant is in the population, and much more. Everything runs from your own computer.
What You Need
- A genome sequencing file in VCF format. Most whole-genome and whole-exome sequencing services provide this. Note that consumer ancestry services like 23andMe provide a different format (raw SNP array data) that is not compatible. You need whole-genome or whole-exome sequencing data, not ancestry chip data.
- A Dnaerys variant store installed on your computer — this is the local database that stores and indexes your genome so it can be queried (install guide)
- Java 21 — needed to run the server (download)
- An AI assistant that supports MCP — such as Claude Desktop, Goose, or another MCP-compatible client
Installation
1. Build a single uber-jar
./mvnw clean package -DskipTests -Dquarkus.package.jar.type=uber-jar
-
This builds
target/mygenome-runner.jar -
Alternatively - download
mygenome-runner.jarfrom the latest release
2. Make sure your Dnaerys variant store is running - it should be available at http://localhost:8001
3. Configure & start AI assistant
Claude Desktop
Open the Claude Desktop configuration file — on macOS it is at ~/Library/Application Support/Claude/claude_desktop_config.json,
on Windows at %APPDATA%\Claude\claude_desktop_config.json, on Linux at ~/.config/Claude/claude_desktop_config.json —
and add the following inside the mcpServers section (create the file if it does not exist):
{
"mcpServers": {
"MyGenome": {
"command": "java",
"args": ["-jar", "/path/to/mygenome-runner.jar"]
}
}
}
-
Replace
/path/to/with the actual location of the jar file on your computer -
NB: port 9000 on localhost should be available, otherwise Claude Desktop will fail to start the service
-
Restart Claude Desktop - you're ready to go
Goose
-
Start the service:
java -jar /path/to/mygenome-runner.jar -
Start Goose, add a new extension with Type "Streamable HTTP" and Endpoint "http://localhost:9000/mcp"
Privacy Policy
MyGenome is a read-only bridge between your local genome database and your AI assistant. It does not collect any personal information. It does not use cookies or tracking. It does not record your conversations. It does not send any data to the internet.
Your Dnaerys variant store runs entirely on your machine and does not phone home.
The privacy risk comes from the AI assistant you choose to use. If you use a cloud-based AI, your genomic data will be transmitted to that provider as part of your conversation. If privacy is your priority, use a locally deployed AI model — your data shouldn't leave your computer.
Related Projects
For those of you involved in medical and genomic research, here is a sister project: onekgpd. Feel free to use it with your clinical cohorts and local LLMs.
Support
- Issues and questions: https://github.com/dnaerys/mygenome/issues
- Email: mcp@dnaerys.org
License
Apache License 2.0 — see LICENSE for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。