Connected AI · across all three

Use the AI you want. Bring the important context with you.

Your chats and agents can live anywhere. HeroMind gives the tools you connect access to the durable context that matters, and a controlled way to return actions, knowledge and decisions. Setup is required; external tools cannot reach HeroMind until you authorize a connection, and built-in AI works without one.

Full web appFree during betaAccess by invitation

Your work can happen anywhere. HeroMind remembers what matters.

The boundary

Your tools stay your tools. What matters comes back.

HeroMind is not the place where all your work happens. It is the place the important outputs return to, and the place a connected tool can ask about you before it starts.

Outside the boundary · yours
AI you already useClaude, ChatGPT, Claude Code, and other compatible tools
Where the work livesEmail, calendar, browser, documents, code
Externally configured agentsOnes you set up and run yourself. HeroMind does not ship them.
HeroMind
Planner Today
Library Over time
Your record the decisions, preferences and mental models you choose to carry forward
Goals & Direction Your direction
What crosses it
←HeroMind → connected tool: relevant personal context and selected knowledge, when the tool asks.
→External agent → HeroMind: sourced tasks, Library sources, notes, and contributions to your record.
Both arrows require a connection you set up. Nothing crosses the boundary on its own.
Context out

A new topic starts by asking what it should know.

The important thing an AI learns about you should not be trapped in one conversation. When a connected tool begins a new topic, it can ask HeroMind for a bounded pack of context, sized to the question.

  • Who you are, in your own words.
  • Standing preferences: how you like things done.
  • Decisions and stances that bear on the topic.
  • Passages from your record and selected Library knowledge.
  • Recent events and today's plate, when they are relevant.

The pack includes only sources and pages you have left visible, and never a suggestion you accepted privately. You choose visibility when accepting a record contribution; existing record entries don't have a visibility toggle yet.

Context pack · “planning a weekend photo shoot”
Product concept, demo content
Who you're working with
Maya. Product manager, learning photography on weekends. Prefers short answers and concrete next steps.
Standing preferences
Plans as checklists, not essays. Metric units. No bullet lists in chat replies.
Decisions that bear on this
Shoot RAW+JPEG until editing is trusted (Sep 27). Stay in aperture-priority until the meter is predictable.
From your record
Goal: a full set in manual mode, October. Preference: aperture-priority for now.
On your plate
Saturday: golden hour 6:41pm. Pack the ND filter.
About 900 characters. Nothing marked private. Sources named on request.
Contributions back

Write back only what should outlive the conversation.

A connected tool can send HeroMind the durable result of its work, not the work itself. The draft, the analysis, and the long chat stay where they are useful. What comes back is small and traceable.

  • A task, with the originating email, run, or conversation attached.
  • A source for the Library, marked with which agent filed it.
  • A daily note from an agent, kept apart from your own notes.
  • A decision, preference, or mental model it noticed, offered to your record for review.

A selected conversation or a useful meeting can be worth keeping too. The rule is about purpose, not file type: import what should shape the future, not everything by default.

Arrives in HeroMind
Task · Work · Thursday
Follow up on the renewal quote for the analytics tool
Source: email thread, filed 6:40am
Library source · Photography
“Golden-hour metering, explained” (demo article)
Filed by your research agent
Proposed for your record
Decision: switch to back-button focus
Waiting on the Desk, your review queue
Not here: the 40-message chat that produced the decision. It stays in the tool you had it in.
Product concept, demo content.
What you could connect

Four workflows, all configured outside HeroMind.

These are agents and tools you would set up yourself and point at HeroMind. HeroMind does not ship an email agent or a finance agent, and it never touches your inbox or your accounts because you signed up. HeroAgents is one possible external system for running them; it is not required, and it is not what this page is selling.

Email → a sourced task

Externally configured agent workflow
Inbox→your email agent→Planner task

An agent you run reads your mail, spots a commitment, and files a task with the thread attached. In HeroMind you see the task, the day you gave it, and a link back to the originating email.

Research → Library material

Externally configured agent workflow
Web, papers→your research agent→Library source

A research agent files what it found as a source on a topic you follow. Overnight the librarian folds it into the topic's reference page and cites it. You read the page, not the raw pile.

Card review → an item to look at

Externally configured agent workflow
Statement→your card-review agent→a sourced Planner task

If you set up an agent to read your own statements, it files an ordinary Planner task with its source: review the renewal due Friday, look at a charge you did not expect. The statement itself stays outside HeroMind. You set this up; HeroMind never touches your accounts.

A conversation → a decision worth keeping

Externally configured agent workflow
New chat→asks for context→works→proposes a decision

A connected AI starts a new topic already knowing what you decided last week. When the conversation produces a durable decision, a workflow configured to propose files that one line for your review.

Control

You decide how AI participates.

Not everything needs a yes, and not everything gets to skip one. Three modes, plainly:

Arrives directly

Routine contributions you authorized

Tasks, sources, and agent notes from a connection you set up land with their source attached. Nothing can be deleted by an agent.

Submitted for review

Suggested changes to your record

A decision, preference or mental model an AI noticed is filed for your review. Only you, in a session you're running, can have something written directly. Accept, edit, or ignore. Editorial syntheses work the same way.

On your instruction

Direct capture when you ask

“Record that I'm switching to back-button focus” writes it now. You said so, in a session you're running.

Syntheses are proposed by default

The librarian offers an editorial when a set of sources is more useful together. If you want a topic written up automatically, you can switch that topic to automatic; it stays off everywhere else.

Superseded, never erased

Change your mind and the old entry is marked superseded and kept. Keep a source or page private, or accept a suggestion privately, and connected AI cannot read it. Revoke a connection at any time from the token page linked in Settings.

No setup needed

The AI that's already inside works on day one.

Everything on this page so far is about tools you connect. You do not have to connect anything to get help from HeroMind itself.

  • Triage: the morning brain dump becomes tasks, notes, and list items you confirm.
  • Ask: questions across your Library, answered from your own sources with citations.
  • The librarian: compiles your sources into reference pages overnight and proposes syntheses.
  • Coaches and Heroes: AI-generated perspectives you set up and talk to.

No API key of your own. Model calls are made server-side, and the providers involved are named on the data page.

Built in

Works without a connection

Triage · Ask · the librarian · Coaches and Heroes · organizing, synthesis, and questions inside HeroMind.

Connected

Needs setup

Context out to a compatible tool · sourced tasks, sources, notes, and record proposals back · externally configured agents you run yourself.

Connecting a compatible tool

One endpoint. Sign in, approve, done.

HeroMind exposes a single connection point that compatible AI tools can use. You sign in and approve the connection on a consent screen; there are no keys to paste. A connection acts with your access to the tools listed in the docs, reading and writing alike; a read-only mode does not exist yet, so connect tools you trust. Compatible means tools that support the open MCP connector standard, not every AI product.

ChatGPTSettings → Connectors → create a connector with the HeroMind endpoint, then approve the consent screen.
ClaudeSettings → Connectors → add a custom connector with the same endpoint. Works in Claude on the web and desktop.
Claude Code and other MCP clientsAdd the endpoint as an HTTP MCP server and authorize once in the browser.
RenewalA connection stays approved for about a month, then asks you to sign in again. Revoke any connection sooner from the token page linked in Settings.

Tool names, what each one can read and write, and the full setup walkthrough are in the docs. MCP and API docs →

The idea in one line

Chats can stay where they are. The context that matters can come with you.

Take command of your day.

HeroMind is in private beta. Request access and I'll send an invitation when a seat opens.

Full web appFree during betaAccess by invitation
A narrower Planner-only iOS beta runs on TestFlight; invitations mention how to get it.