Genomi MCP
Genomi parses a DNA file on your machine while an agent matches variants against ClinVar, GWAS and drug-response data.
- 493 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The runtime handles personal genetic data and local libraries.
- Health conclusions need confirmation from a clinician.
Install
- Works in
Manual install
Install and configure Genomi by following the instructions here: https://raw.githubusercontent.com/exon-research/genomi/master/INSTALL_FOR_AGENTS.md
A prompt for an agent that performs the install from the guide.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The runtime handles personal genetic data and local libraries.
- Health conclusions need confirmation from a clinician.
This is third-party code. Review the repository files before installing.
What it does
Genomi is an open-source runtime that turns an AI agent into a personal genetics helper. A genome file (VCF, gVCF or an export from 23andMe, AncestryDNA, MyHeritage and others) is parsed locally into an Active Genome Index, and the agent queries only the variants it needs. Public sources sit beside the index: ClinVar, GWAS, pharmacogenomics, ancestry and polygenic scores. A journal records what the agent explored and the evidence behind each conclusion. It works with Claude Code, Codex, OpenClaw, Hermes and any MCP host. Private reads need approval in the current session.
Who it is for. People holding a DNA file from 23andMe or a sequencer who want to explore it with an agent and keep the raw data at home.
Good fit when
- The question is how your body responds to a drug or caffeine, and the answer should rest on your own variants.
- A 23andMe export or VCF sits on disk and should be opened to an agent without uploading it to a third-party service.
- An investigation spans several sessions and the agent needs a journal of what was already explored.
- An agent's risk answer should be checked against ClinVar records and population frequencies.
Not a fit when
- You need a diagnosis or treatment decision: the project positions itself as a research helper.
- There is no disk space for the public libraries: the full install is large.
- A genome file cannot be opened even to a local agent backed by a cloud model.
Example request
Open my 23andMe file and tell me whether I metabolise caffeine fast or slow, with source references.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
The project is at version 0.1.0, and GenomiLab (the disease-research workspace) is a developer preview. It is a research helper, not a diagnostic or treatment tool: confirm findings with a clinician. The full library install (`everything`) pulls a lot of data; smaller sets such as `common-questions` save space. Answers depend on the connected LLM, so the model sees query results even when the raw file stays on your machine.
How to disable. Remove the `genomi` entry from the MCP host configuration, and the `~/.genomi` directory if you no longer need the index, journal and libraries. Genomi does not modify the genome file itself.
FAQ
Which files does Genomi accept?
VCF and gVCF, raw exports from 23andMe, AncestryDNA, MyHeritage, FamilyTreeDNA and Living DNA, plus VCF files from Nebula, Dante Labs and Sequencing.com. Paired FASTQ files are aligned locally.
Does the genome leave my machine?
The raw file stays on your computer and the index is built locally. Public lookups go out by chosen targets: an rsID, gene, drug or condition.
How do I try it without my own DNA?
Take a public export from a participant of the Harvard Medical School Personal Genome Project and point the agent at it; the README calls this the cleanest way to try it.
How do I call Genomi in Codex?
Use `$genomi` instead of `/genomi`. Other hosts connect `genomi serve` over stdio.
Is a Genomi result a diagnosis?
No, the project calls itself a research helper. Findings and clinical implications need confirmation from a clinician.
Discussion
No questions or reviews yet.
Related
MCP serversA Claude Code plugin and MCP server: over 1000 scientific tools for biology, chemistry and medicine, data-check commands and routing skills
A Google DeepMind skill collection for science tasks: access to biology and chemistry databases, literature search and structural biology
A molecular biology workbench for Codex and Claude Science that opens DNA, RNA and protein sequences, alignments and results.
ClawBio is a library of 110 agent skills for bioinformatics: genomics, pharmacogenomics and population genetics, reproducible.
BioMCP
biomcp
One binary and a single command grammar for 70 biomedical sources: PubMed, ClinVar, ClinicalTrials.gov, OncoKB. MCP mode starts with the serve command
NVIDIA BioNeMo Agent Toolkit: skills for life sciences
NVIDIA BioNeMo Agent Toolkit
NVIDIA skills for biology and chemistry: the agent predicts protein structures, docks molecules and designs proteins
README in short
Data is split into layers: your variants, public evidence, reviewed findings, agent memory and personal context. That lets the agent show what each answer rests on.