Mind to AI

A capture inbox for your AI

Capture now.
Let any AI think later.

Mind to AI is a private inbox for half-formed thoughts. One tap on the iPhone records a voice memo or grabs a photo, a screenshot or a note. The app stores it exactly as captured and hands it, whenever you ask, to Claude, ChatGPT or any tool that speaks MCP. It runs no model of its own, and it never will.

See how it works Connect your AI
  • 0models inside
  • 1place for every capture
  • anyAI on the other end
Inboxtoday
  • Voice memo0:42 · "the pricing idea from the car"
  • Screenshota chart from a newsletter
  • Photowhiteboard after the call
  • Note"ask about the Q4 budget"
Your AIconnected

"Transcribe this morning's memo and turn it into three bullet points for the pricing page."

  • Claude
  • ChatGPT
  • Cursor
  • Claude Code
  • anything with MCP

Three moves. Nothing clever in between.

  1. 01

    Capture

    On the iPhone, an Action Button press or one tap in a Shortcut starts recording. Share a screenshot, snap a photo, type a line. In the browser, drag a file in or talk into the mic. The capture lands in your inbox in the time it takes to lock the phone.

  2. 02

    Store

    The original file, untouched, on a server you control, deduplicated by content and tagged the way you like. Inbox, archive, trash. No transcript, no summary, no "smart" anything is written by the app itself.

  3. 03

    Hand over

    Point any AI at it. A remote MCP server and a plain REST API with an OpenAPI spec expose the same inbox. Claude reads the memo, ChatGPT sorts the screenshots, a script does the rest. The thinking happens on their side, on your terms.

Connect any AI.

Mind to AI does not pick a provider for you. It publishes your inbox over two standard doors and lets you decide who walks through.

  • MCP at mcp.mindtoai.one: works in Claude, Claude Code, Cursor, ChatGPT and every other client that speaks the Model Context Protocol.
  • REST + OpenAPI: the same endpoints for custom GPT actions, function calling, or a twenty-line script.
  • Per-client tokens you create and revoke yourself. Each AI gets its own key; pulling one never touches the others.
  • Nothing is pushed. A capture leaves the server only when a client you connected asks for it.
Add it to an MCP client
{
  "mcpServers": {
    "mindtoai": {
      "url": "https://mcp.mindtoai.one/mcp",
      "headers": {
        "Authorization": "Bearer <your token>"
      }
    }
  }
}

Then: "list what I captured this week", "transcribe the last voice memo", "tag every screenshot with a chart as finance". The client fetches, the model works, the inbox stays as it was.

The contract.

Short, because the app does little on purpose.

What the app does
Records voice, takes and receives photos and screenshots, keeps typed notes.
Stores the original bytes, content-hashed, on a server behind a login only you hold.
Serves the same inbox to the web app, the iPhone and your AI clients.
Lets you mark things processed, tag them, archive them, throw them away.
What it never does
Run a model. There is no transcription, no OCR, no summary, no embedding inside the app.
Hold an AI provider's key. It cannot call Claude or OpenAI even if you asked it to.
Send a capture anywhere on its own. Only a client you connected can pull one.
Change what you captured. Every byte you recorded is the byte you get back.

On the iPhone, two ways in.

A Shortcut comes first: record, snap or pick, upload, done. It needs no App Store and never expires. The native app follows for the parts only native can do: Action Button launch, a lock-screen widget, recording with the screen off, a share sheet for screenshots.

Video capture is built into the pipeline and switched off for now, until large uploads get the treatment they deserve.

In development

Private build. Not in the App Store. This page describes what is being built, and it changes as the build does.

Questions.

Why not just use Voice Memos and Photos?

Because your AI cannot see them, and because they mix your captures with everything else. Mind to AI is one inbox, reachable by any AI you connect, holding only the things you meant to think about later.

Does it transcribe my recordings?

No. The app stores the audio. When you ask Claude, ChatGPT or another connected client about a memo, that client fetches the audio and does the transcribing on its side. The app stays dumb storage on purpose, so it never has to know which AI you prefer.

Where is my data?

On a server the account owner runs, behind Cloudflare Access. Nothing is on a third-party cloud unless you connect one. The privacy page lists every connection the app makes.

Which AI tools work with it?

Anything that speaks MCP: Claude, Claude Code, Cursor, ChatGPT and a growing list. Anything that can call a REST API with a bearer token works too, through the OpenAPI spec.

Can I get it?

Not yet. It is a private build under development. This page will say so when that changes.