Back to browse
GitHub Repository

An open-source agent harness that turns your AI agent into your personal DNA expert

170 starsPython

Genomi – make genome data manageable and queryable by AI agents

by matthewzmd·Jun 3, 2026·2 points·0 comments

AI Analysis

●●●BangerZero to OneBig BrainDark Horse

Local-first AI for DNA queries when sending genetic data to cloud APIs feels genuinely risky.

Strengths
  • Solves the privacy paradox of AI genomics by keeping raw DNA files entirely on-device
  • MCP integration works with Claude Code, Codex, and any MCP-capable agent host
  • Active Genome Index converts overwhelming variant data into queryable evidence-backed answers
Weaknesses
  • Requires users to already have raw DNA files from 23andMe, Ancestry, or similar services
  • Medical interpretation risks need clearer disclaimers about non-diagnostic use
Category
Target Audience

Biohackers and individuals with raw DNA data from testing services

Similar To

Prometheus · Nebula Genomics

Post Description

Hey HN, I’m Matthew, one of the people building Genomi.

Many people have done a DNA test or sequencing in the past, but the results stay virtually dead on our computer, maybe we saw a report that came with it, but does anyone still remember what it said? I don't. What bothers me even more is that the world of DNA science actually evolves quite quickly, there are many new findings published every year and that may or may not be relevant to your DNA, but your old report, which I spent a fair amount on, will not have it! (though it has gotten much much cheaper lately)

I feel AI agent is the perfect tool to help me query and understand my own DNA, but I am immediately hesitant to feed my own file to an agent, not only the security risk of sending my own DNA file to them, but it simply doesn't work, the massive nature of the file will overflow the context window with excess data that gives me near to no value. So we built Genomi: an open-source local genomics harness for AI agents.

Genomi parses a VCF/gVCF or supported consumer DNA export, such as 23andMe or AncestryDNA, into a local SQLite index called the Active Genome Index (AGI, heh). The agent does not need the whole genome in context. Instead, it calls tools against the local index and public evidence sources as it sees fit. Today it exposes 80+ evidence-focused tools across genomics sources including variant lookup, gene/disease evidence, pharmacogenomics, GWAS/PRS context, population context, sequence/region utilities, etc.

The project is at an early stage and I welcome any feedback, please let me know what you guys think!

Similar Projects

AI/ML●●●Banger

Build queryable packs for AI agents from videos, podcasts, and files

Cross-agent packs work with Claude Code, Cursor, and Ollama — build once, query anywhere.

Big BrainSolve My ProblemDark Horse
buralog
102mo ago