● Your choice of AI

Use Selfscope with your AI

Ask Claude, ChatGPT or another AI about your own lab results, timeline and records. You choose what it sees, and your data is only ever decrypted on your own devices.

Two ways to use your AI

MCP (the Model Context Protocol) is the standard way AI apps connect to tools and data. The Selfscope MCP server runs on your own computer: it downloads your data still encrypted, decrypts it there with your key, and answers your AI. Selfscope's servers never see it readable. It can't change anything, and you can switch it off at any time.

Set up the MCP server (once)

For founding testers. You need an activated invite key, Selfscope Cloud turned on, and Node.js 22 or newer on your computer. The @selfscope/mcp package is being published; if the command below can't find it yet, write to hello@selfscopehealth.com and we'll send you early access.

1. Create an agent key. In Selfscope, open the account menu, choose Cloud sync, then under AI agents choose Create agent key. Copy it: it's shown only once.
2. Connect it to your data. In a terminal, run:

   npx -y @selfscope/mcp setup

   Choose 1 (Selfscope Cloud), paste your agent key, pick your account and enter your passphrase. Your passphrase is used once, on your computer, and isn't saved or sent anywhere.
3. Add it to your AI app using the steps for your app below, then restart the app.

Setup saves your connection, including your account's data key, in ~/.selfscope/mcp.json. Anyone who can read that file can read your data, so keep your computer account private and use disk encryption if you can. Prefer to stay offline? Choose 2 in setup and point it at an encrypted backup file (account menu, Back up); your AI then sees the data as of that backup.

Claude Desktop

1. Open Claude Desktop's settings, go to Developer and choose Edit Config. This opens claude_desktop_config.json (on Windows in %APPDATA%\Claude\, on a Mac in ~/Library/Application Support/Claude/).
2. Add Selfscope, keeping any servers already there:

   {
     "mcpServers": {
       "selfscope": { "command": "npx", "args": ["-y", "@selfscope/mcp"] }
     }
   }

3. Quit and reopen Claude Desktop. Selfscope appears in the tools menu of a new chat.

Claude Code

claude mcp add selfscope -- npx -y @selfscope/mcp

Then start Claude Code and type /mcp to check that selfscope is connected.

ChatGPT, Claude.ai and phone apps

These run in the cloud, not on your computer, so they can't use a local MCP server. Use Copy for AI instead: in Selfscope, open the account menu, choose Copy for AI, choose what to include, then copy the text into your chat (or download it as a file and attach it). Your name and date of birth are never included.

A direct connection for web and phone AI apps may come later. It would mean our servers reading the data you choose to share, so it would be strictly opt-in. We'll ask founding testers before building it.

Cursor

Open Cursor's settings, go to MCP and add a server, or edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "selfscope": { "command": "npx", "args": ["-y", "@selfscope/mcp"] }
  }
}

VS Code (GitHub Copilot)

Run MCP: Add Server from the Command Palette, or add this to your user or workspace mcp.json, then use Copilot Chat in Agent mode:

{
  "servers": {
    "selfscope": { "type": "stdio", "command": "npx", "args": ["-y", "@selfscope/mcp"] }
  }
}

Gemini CLI

Add Selfscope to ~/.gemini/settings.json:

{
  "mcpServers": {
    "selfscope": { "command": "npx", "args": ["-y", "@selfscope/mcp"] }
  }
}

OpenAI Codex CLI

Add Selfscope to ~/.codex/config.toml:

[mcp_servers.selfscope]
command = "npx"
args = ["-y", "@selfscope/mcp"]

Other MCP apps

Any app that supports local (stdio) MCP servers works. Add a server with the command npx and the arguments -y @selfscope/mcp; most apps use the same mcpServers format as Claude Desktop above.

What to ask

Your AI can see an overview, every lab marker and its history, the changes you've logged, readings before and after each change, and your records. Try:

• "Give me an overview of my Selfscope data. What stands out?"
• "Which of my lab results are outside the lab range, and how have they moved over time?"
• "What changed in my results after I started magnesium, and what else did I change in the same period?"
• "Summarise my medical records from the last year for a doctor's appointment."

Selfscope shows what changed over time, not what caused it. Your AI can help you spot patterns and prepare questions, but it isn't medical advice: talk to a clinician about your results.

Turning it off

• Stop one AI app: in Selfscope, open Cloud sync, then AI agents, and choose Revoke next to its key. It can't read anything after that.
• Remove it from this computer: run npx -y @selfscope/mcp forget to delete the saved connection and key, and remove the server from your AI app's settings.
• Check what it's connected to: npx -y @selfscope/mcp status.

Help

"Not set up yet"

Run the setup command in a terminal first (step 2 above), then restart your AI app.

"This agent key was revoked…"

The key was revoked, or your invite key is no longer active. Create a new agent key in Selfscope and run setup again.

"That passphrase doesn't open this account"

Use the passphrase you use in Selfscope for that account. If you changed it recently, use the new one.

"No accounts found"

Turn on Cloud sync in Selfscope first, on the device where your account is, then run setup again.

Your AI doesn't see new data

The server checks for new data about once a minute. Make sure Cloud sync is on and has synced on the device where you made the change.

What does the AI app do with my data?

Whatever your AI app sends to its model is covered by that app's own privacy terms. Selfscope only decides what the server can read: your own data, read-only.

Still stuck? Write to hello@selfscopehealth.com, and please don't include any of your health data.