What is MCP, and how does your AI read your genome?

Ask any AI assistant "how much coffee is too much?" and you get the answer for the average human. The average human is not you. Your assistant knows nothing about your caffeine clearance, your chronotype, or your recovery genetics, so it hedges, and hedged advice is hard to act on.

The fix is not a smarter chatbot. It is giving the assistant you already use a way to look up data about you. The plumbing for that is called MCP.

What MCP is

MCP stands for Model Context Protocol. It is an open standard, introduced by Anthropic in late 2024 and since adopted across the industry, including by OpenAI and Google, that defines how an AI assistant connects to outside data and tools.

A good mental model is a USB port. Any data source can expose a standard plug, any compatible assistant can use it, and neither side needs a custom integration for the other. One standard, everything connects.

How Helisoma uses it

Helisoma runs an MCP server whose whole job is to serve one thing: your genome report. Claude supports these connectors natively: one button on your account page opens Claude's "add connector" dialog, you approve read-only access, and that is the entire setup.

ChatGPT and Gemini do not take personal connectors like this yet, so there the same idea runs through a file instead: you attach your report PDF once, to a ChatGPT Project or a Gemini Gem, along with our ready-made coaching prompt. Your report does not change after analysis, so the attached copy never goes stale. Either way it is one setup, done once.

From then on, whenever a conversation touches sleep, caffeine, nutrition, training, stress, or anything else where genetics matters, your assistant reads your report, live through the connection in Claude, from the pinned copy in ChatGPT or Gemini: every trait, your genotype at each marker, the plain-English interpretation, and the linked studies. It reads, then answers you specifically.

In plain words

Think of it as giving your assistant a library card that can check out exactly one book: your report.

The assistant does not get your raw DNA file, does not root around in your account, and cannot check out anything else. When you ask something relevant, it opens the book, reads the pages it needs, and answers. When you revoke the card, the library closes.

Connect once, then just ask

Here is the part that changes how this feels day to day: after the one-time setup, you can forget the file, forget the report, forget that any setup happened. Nothing to re-find, re-upload, or re-explain. You just ask questions the way you already talk to your assistant, and add "for my DNA" when you want the answer to be about you:

• "I have a hard workout Thursday and a late flight Friday. Plan my caffeine around that, given my genetics."
• "From my DNA's point of view, am I better off training in the morning or the evening?"
• "I keep waking up at 3 a.m. What in my report is worth looking at, and what should I try first?"
• "Build me a grocery list for the week that fits my nutrition markers."

In Claude the assistant fetches your report itself each time; in ChatGPT and Gemini it reads the copy pinned to your project or Gem. Either way the answers stay grounded in your data even in a brand-new conversation, months later, on your phone, in the middle of something else. That is the point: your genome becomes something your AI just knows, not a file you have to go find.

The honest part

Three things worth saying plainly. First, answer quality depends on the assistant you connect; MCP delivers the data, it does not think for the model. Second, when you connect, your report goes to the AI platform you chose and is handled under that platform's own privacy policy, so read it, and disconnect any time; the connection exists only at your direction. Third, the report is educational: population-level leans with linked evidence, not a diagnosis, and anything clinical belongs with a doctor.

Why this costs $49 once, not $19 a month

Most "AI health coach" products charge a subscription because they run the AI. Helisoma does not sit in the middle. The analysis is done once, the report lives in your account, and the conversation runs on the AI subscription you already pay for. There is nothing for us to meter, so there is nothing to subscribe to.

Upload your raw DNA file, get your report, connect your assistant, and ask it something you have always wondered about your own body. That is the product.