H E L M

Plain language, on purpose

Privacy

Last updated August 17, 2026

Helm’s whole idea is that your life lives in one Markdown file you own. That shapes everything here: we collect little, we can show you exactly where every byte lives, and you can take the file and leave whenever you want.

The short version:we store your brain in a private GitHub repo, send captures through an AI model to file them, and run basic analytics on the public site. We don’t sell data, don’t show ads, and don’t train models on your content. Download your file anytime; email us and we delete everything.

What we collect

  • Sign-in identity. When you continue with Google we receive your name, email address, and profile photo — nothing else. We never see your inbox, contacts, or calendar. Your session is a signed cookie on your device.
  • Your brain.Everything you put in Helm — tasks, captures, notes, photos you extract from, links you save — becomes part of your Markdown file. It’s the product; of course we store it.
  • Feedback and waitlist notes. If you send feedback, we keep what you wrote (and your email if you include it).
  • Basic analytics.On the public site we count visits and funnel steps (Google Analytics and Vercel Analytics) to see what’s working. No ad trackers, no cross-site anything.

Where your brain lives

Your file is stored as a private GitHub repository. Every change is a git commit, so you get full history for free — nothing is ever silently lost.

Honest part: because we host that repo for you, Helm holds admin access to it. We use that access only to run the service — reading and writing your file when you use the app. We don’t browse, analyze, or share your brain, and we’d rather build features than look at your grocery list.

AI processing

When you capture something, an AI model (Anthropic’s Claude, via Vercel’s AI Gateway) reads it to decide where it files — that’s how “milk” lands in To Get and a recipe link lands in Recipes. Photo captures are read by a vision model to extract the items. Links you save are fetched by our server to grab the title and thumbnail.

These requests are processed with zero data retentionat the gateway, and Anthropic does not train on API traffic. Your captures pass through the model; they don’t stay there and they never teach it anything.

Who touches your data

We run on a small set of providers, each doing one job: GitHub (stores your repo), Google (sign-in only), Vercel (hosting, analytics, AI Gateway), Anthropic (the AI model), and Resend(relays feedback emails to us). That’s the whole list. No data brokers, no advertisers, ever.

Your file, your rights

  • Export:download your complete brain from the app anytime. It’s plain Markdown — readable by you, your editor, and any AI you point at it. No lock-in is the point.
  • Delete: email us and we delete your repo — content, history, account, everything. Gone means gone.
  • Leave: the free path never required an account. If you built a brain and downloaded it, we hold nothing of yours at all.

The boring-but-important bits

Helm isn’t for children under 13, and we don’t knowingly collect their data. If we ever materially change this policy, we’ll say so here with a new date up top — and if a change affects your brain’s privacy, we’ll tell you before it takes effect, not after.

Questions, exports, deletions: hello@madronaproduct.com. Helm is built by Madrona Product Studio.