Gethrsgethrs

AI as support: why Gethrs talks to Claude

Written by the people who build Gethrs. It says what we do with artificial intelligence, what we deliberately do not do, and how to connect Claude to your notes if you want to. No promises about the future: only what exists today.

The Gethrs home page says "no AI in the way". Some people asked whether we are against artificial intelligence. We are not. We are against one specific idea of it: the kind that sits between you and what you are writing. The suggestion that pops up while you think, the summary you did not ask for, the "improve" button next to every sentence. In an app for taking notes, that stuff does not help: it distracts, and slowly convinces you that writing is a job to delegate.

We believe in something else: AI as support. A tool you call, when you need it, to do something you decided. That is why the Gethrs editor has no intelligence inside, and that is why Gethrs can be connected to Claude.

What we deliberately do not do

Nothing in the editor writes for you. No autocomplete, no rewriting, no summary that starts on its own. The text you see is the text you wrote, and it stays that way until you change it.

The features that go through a model do exist, but they are all off until you switch them on, and each does one thing you asked for:

  • Reading the text in a photo (a whiteboard, a handwritten note): the image goes to Azure AI Vision and the text comes back, under the photo. You switch it on in Settings.
  • Recording a call and transcribing it: the audio goes to Azure Speech in chunks and the text lands at the bottom of the note. It starts when you press the button.
  • The connection with Claude, described below. You create it in Settings and revoke it there.

When you do not use them, no external service sees your notes. We say so in the privacy policy too, because that is the point.

What Claude does, if you connect it

Gethrs exposes an MCP server (Model Context Protocol), the standard Claude and other assistants use to talk to applications. Once connected, Claude can do seven things in your notes, and only those:

  • Create a note, with a title, simple markdown text, a notebook and tags.
  • Append text to the end of a note that already exists.
  • Search your notes, with the same index the app uses: titles, text, even the text inside attached PDFs and photos.
  • Read a whole note, with its tags, notebook and tasks.
  • Create a task, with a due date and priority, inside a note or on its own.
  • List tasks: open, today, overdue or done.
  • Tick a task off, updating the line in the note as well.

A note made by Claude is a note like any other: it goes through the same functions of the app, lands in the same index, shows up in the same Library. There is no separate "AI corner".

A concrete example. You finish a call with a client, open Claude and say: "make a note in Gethrs with the points of this call and add a task for Friday: send the quote". Claude creates the note in the notebook you name, with the task inside, and the task shows up in your Tasks view with the right due date. Or: "search my notes for what we decided about the logo in March", and Claude searches and brings back the excerpt, with a link to the note.

Why Claude does nothing on its own

The connection is a token tied to your account: whoever has it can write in your notes, so you create it, you see it once, and you revoke it whenever you want from Settings → Claude. On the same page you see when it was last used.

Claude acts only when you ask, in its own window. There is no process running at night, no AI "watching" your notes to suggest things. The MCP server answers a request from Claude, does what was asked, and stops. If one day you no longer want it, you revoke the token and the connection no longer exists.

How to connect, in two minutes

On claude.ai or Claude Desktop, with no token to copy: Settings → Connectors → "Add custom connector", paste the address https://gethrs.app/api/mcp, press "Connect", sign in with your Google account and say "Allow". It is OAuth, all on our domain: Claude gets a token of its own, which appears in Settings → Claude as "Connector · Claude" and can be revoked from there.

In Claude Code, in the terminal: create a token from Settings → Claude and then

`` claude mcp add --transport http gethrs https://gethrs.app/api/mcp --header "Authorization: Bearer <your token>" ``

Then ask Claude whatever you need: "create a note in Gethrs with…", "add a task for Monday…", "what did I write about…".

Where we want to go

We do not want an app that thinks for you. We want an app where the notes are yours, readable, exportable, and where the smart tools are within reach, not in the way. Claude through MCP is the first step of that idea. If you try it and something is off, write to us from the "Feedback" button at the bottom right: a person answers.

Join the beta: instant access

Start from scratch or bring your notes along, and tell me what breaks. I reply personally within 48 hours.