On your phone, and Settings

HeroMind on iOS

HeroMind ships as a native iOS app with the same account and data as the web. Sign in with the native Apple sheet (Face ID, no browser round trip) and stay signed in: the session survives backgrounding, restarts, and days away.

Capture is the point on the phone. Add a task, flag it for today, drop a link into the Library, dictate a brain dump, save a page from Safari's share sheet. The fuller desktop layout is waiting when you sit down, and everything you did on the phone is already there.

What is on the phone that is easy to miss:

  • Back. A back arrow in the top bar, plus edge-swipe, so a deep link from the Desk to an editorial to a source can always be unwound.
  • Library search in the masthead, under the tabs.
  • + Add source with all six tabs, including Research (shown as a magnifying glass at phone width).
  • Print still works; it opens the print sheet as a PDF you can share.

If a page ever sits on "Loading", the app retries on its own and reloads once. If that fails you get a Reload button and a Go to today button rather than a blank screen. Killing the app is never required.

The browser extension

The Chrome extension saves the page you are on into the Library, and optionally onto today's focus list, in one click. Pair it once from Settings; it uses your signed-in session, no separate password.

Settings

  • Theme. Light, dark, or follow the system.
  • Personalize my assistant. How you plan, how proactive the assistant should be, brief length, quiet hours, and the patterns it has drawn from your task history. The assistant chapter covers each control.
  • AI activity. Every model call HeroMind made for you since the start of the month, by feature, with tokens and cost. HeroMind pays for these; you never need an API key. If a model is unavailable the call falls back to a second model on its own, so a failure here means every option failed.
  • Library. Editorial proposals on or off, the cool-down between them, and whether topics set to Automatic get editorials written without approval.
  • Agent access (MCP). Create a token per agent, see when each was last used, and revoke any of them. Interactive clients (ChatGPT, Claude, Claude Code) connect with a browser approval instead and appear here too.
  • Calendar. Connect external calendars (read-only) so they appear on the Time Blocks tab.
  • Export and delete. Your tasks, notes, journal, Library, and record export as Markdown. Delete everything, anytime.

Your data

Brain dumps and journal entries are among the most personal data anywhere. Discard a triage session and the raw text is permanently deleted. Every AI suggestion is read-only until you confirm it. What the advisor learns about how you work is a mirror, not surveillance, and you hold the eraser. No selling your data, no training on it without your say-so.

The same holds for the agents you connect. Each has a token you can revoke, every call is logged, and nothing an agent proposes becomes true about you until you accept it on the Desk.

Updated 2026-10-11 · All chapters · API reference