Indaga speaks MCP, the open protocol AI assistants use to reach your data. Point Claude, ChatGPT, Grok or any other assistant at your record and ask in your own words. The assistant does the talking; Indaga supplies the evidence and its limits.
One tap. Set it up once and it follows your Claude account — the web, the desktop app and the Claude app on your phone.
1
Add Indaga to Claude. The button opens Claude’s connector setup with the name and address already filled in. Press Add.
2
Sign in to Indaga and approve. Claude sends you here, you say yes to this one connection, and Claude is issued its own credential — one you can revoke later without touching anything else.
3
Ask. Start a new chat and try “What does Indaga know about my sleep?”
Add Indaga to ClaudeClaude should detect that Indaga needs a sign-in (“Authentication: Always required”). If it shows “None”, switch it — “None” tells Claude never to ask for a credential, and nothing will load.
On a phone, the button opens claude.ai in your browser. Once you approve, the connection is already there in the Claude app — nothing to set up on the phone itself.
ChatGPT, Grok and everything else
Any MCP client connects with the same address and the same sign-in. Copy it, paste it where your assistant asks for a server, and choose OAuth when it asks how to authenticate — you sign in to Indaga and approve, exactly as with Claude.
https://app.indaga.ai/v1/mcp
ChatGPT
Settings → Apps → Advanced → turn on Developer mode
Apps & connectors → Create connector → paste the address
Authentication: OAuth → sign in and approve
Developer mode is available on Plus, Pro, Business and Enterprise plans, from the web app. A connector you create there is then available in the ChatGPT apps.
Grok
grok.com/connectors → New connector → Custom
Paste the address and keep what Grok discovers about it
Sign in and approve
Grok reads Indaga’s sign-in details from the address itself; there is nothing to type in besides the URL.
Desktop tools
Claude Code, Codex, Gemini CLI and Antigravity take a credential in a header instead of signing in. Make one here and paste the config it gives you.
Sign in to download the skills bundle for these tools.
Things to ask
Every answer arrives with its evidence and its limits attached, and the AI is told to show them. Your AI is told which of these your record can answer today, and which it cannot yet — it will say so rather than guess.
What changed this week?Draws on two weeks of wearable data.
Am I recovered enough to train hard today?Draws on HRV and breathing rate.
How regular has my sleep been?Draws on seven nights of sleep.
When is my biological midnight and best training window?Draws on two weeks of heart rate.
How fit am I for my age?Draws on a VO₂max reading and your age.
What does my DNA say about caffeine, lactose, folate?Draws on your genome.
Show me my genome report.Draws on your genome.
What is my cholesterol?Draws on a blood test you uploaded.
What your AI is told
Before your first question, your AI asks Indaga for a briefing about your record — and for the rules it has to answer under. Three things, and nothing else.
The briefing. What your record holds — how many nights of sleep, which years of heart rate, whether a genome or a blood test is on file — in counts and dates. Never a reading: it says a blood test exists, not what the blood test said.
The rules. A value that was never measured is unknown, never normal. A value that is old is quoted with the date it was taken. A percentile, a predictor score or a “Pathogenic” label is not a diagnosis.
The boundary. It reads your record and logs what you tell it, and it can take back what it logged itself. It cannot delete, export or change your settings — the full list is below.
A connected AI can
Read your record — wearables, labs, sleep, genome and the readings derived from them.
Log what you tell it: meals, supplements you took, a supplement plan, a daily target, notes, tags and journal entries.
Propose blood test results it read from a report, for you to confirm in the Indaga app.
Register the routines it runs on its own schedule, which you can stop.
Take back what it logged itself, with a reason.
It cannot
Delete anything, or take back what it did not log.
Store a number it estimated.
Confirm a blood test result for you.
Export your record or start a download.
Spawn a job or change your settings.
Reach any other AI’s credential — revoking one leaves the rest alone.
Answering your questions means sending parts of your health record — including genetic and health data — to whichever AI provider you choose. They handle it under their terms, not Indaga’s, and most process data outside the EU. You will see this again on the approval screen, with the provider named, before anything is shared.