MCP Server

GeneOps speaks MCP (Model Context Protocol) — the open standard for connecting AI agents to external data. Hook up any compatible agent and it gets structured access to your genomics results, action items, and the full research database.

Don’t have an MCP agent yet?

Any agent that speaks MCP will work. Two good options:

Claude Code by Anthropic

Terminal agent with native MCP support. Great for in-depth research sessions.

npm install -g @anthropic-ai/claude-code
OpenCode open source

Open-source terminal agent. Bring any model — including local ones via Ollama.

curl -fsSL https://opencode.ai/install | bash

Also works with ChatGPT desktop, Cursor, Windsurf, Cline, and any other MCP client.

Get your API key

Your API key authenticates your agent with GeneOps.

Go to Account to generate your API key.

Connect your agent

Generate your API key above and it is filled in here.

Paste this into your agent — Claude Code, Codex, Cursor, OpenCode or any other that supports MCP — and it sets up the connection for you: 

Add this MCP server to my configuration:

{
  "mcpServers": {
    "geneops": {
      "type": "http",
      "url": "https://api.geneops.ai/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Or, in Claude Code, run this in a terminal: 

claude mcp add --scope user --transport http geneops https://api.geneops.ai/mcp --header "Authorization: Bearer YOUR_API_KEY"

Available tools

Your agent gets 6 tools — the same data our built-in agent uses, which has every tool except the ones marked MCP only.

get_my_dashboard

Your results by category, or the list of your matched results in one category, with each one's concern flag.

get_my_actionables

All your resolved action items. Filter by type (supplement, diet, lifestyle, monitoring, avoidance) or health category, or open one by id.

get_my_genotype

Deep dive on any single variant by rsID. Returns your alleles, interpretation, concern level, and personalized action items.

knowledge_base

Browse or search the full research database by gene, rsID or keyword. Categories, SNPs, articles, population frequencies and genotype interpretations.

get_my_documents

Your saved documents (plans, protocols and notes saved from earlier conversations, and files you uploaded), or one read by document_id.

list_shared_profiles MCP only

The profiles other people have shared with you, or you with them. Pass a name as for_user to the dashboard, actionables and genotype tools to read a shared profile.

What this gives you

Cross-reference with your own documents. Point your agent at lab results, doctor’s notes, or research papers on your machine.

Choose your own model. Run Llama or Mistral locally, or use Claude, GPT, Gemini. MCP is model-agnostic.

Long-running health research. Query your genotype data, search PubMed, read new studies, keep notes across sessions.

Try asking

“Give me an overview of my results and the most important findings.”

“My homocysteine came back at 14 μmol/L. Is that related to any of my variants?”

“Cross-reference my pharmacogenomics profile with my current medications.”

“Build me a supplement stack based on my actual genetic needs.”

A note on privacy

The MCP server returns your structured genetic data to whatever agent you connect. If you pair it with a local model via Ollama, your data never leaves your machine — only the tool calls hit our server. We don’t log tool call contents.

If you use a cloud model, your data is governed by that provider’s privacy policy. Either way — you choose.

Don’t need your own agent?

GeneOps has a built-in AI agent, and you choose the model for each question. It has access to the same tools and data — no setup, no API key, works in your browser.

The trade-off: with MCP you choose the model, can use local models for full privacy, and can cross-reference your own files (lab results, medical records, research papers). The built-in agent is simpler but runs on our infrastructure and your queries go to the provider of the model you choose.

Try the built-in agent