Writing an idea

Chat apps

Add Walden to Claude, ChatGPT or another chat app that takes connectors, and sign in to it as yourself. Then the chat can read your ideas when you ask about them, start a new idea from what you’ve been talking through, or add to one, all as drafts you keep or discard.

What it’s for

A lot of thinking happens in a chat before it’s an idea. Talk it through for as long as you like, with no Walden in the conversation. When it’s taken shape, say so, and the chat puts it in Walden as yours: your own words, quoted as you wrote them, open the idea’s chat, and whatever the app drafted of the Problem, the Approach, the unknowns and what you’re not doing lands on the page with a dashed edge, counted for nothing until you keep it. See Drafts.

The other way round works too. Ask the chat what’s in Walden, or what an idea says, and it reads the idea in your words and answers from that.

Before anything goes

The chat shows you what will go first: the title, the words of yours it will send, and each part it drafted. It sends only when you say that’s right. If it isn’t, say so and it drafts again. The same when it adds to an idea: it reads the idea first, then shows you what it would add.

Add it in Claude

This is the path checked end to end.

  1. In Walden, go to Settings → Connections → Chat apps and copy the Address.
  2. In Claude, open Customize → Connectors, press + and choose Add custom connector.
  3. Paste the address and press Add. Leave the advanced settings alone: there’s no client id or secret to give.
  4. Press Connect. A sign-in page opens: sign in to Walden with the same email you use here.

In a Team or Enterprise organization, an owner adds the connector under the organization’s settings first, and then each person connects to it from their own Connectors page.

In ChatGPT

ChatGPT takes custom connectors on the plans that allow them with write actions: a Business, Enterprise or Edu workspace, with developer mode turned on by its administrator. Where that’s the case, add Walden the same way, with the same address from Settings → Connections → Chat apps, choose OAuth when it asks how to sign in, and sign in with your Walden email. Where developer mode lives, and whether a personal plan has it, differs by plan and has moved during 2026: follow ChatGPT’s own help for your plan.

What it can and can’t do

It can

  • Read the ideas in every workspace you belong to, and what earlier ideas there decided.
  • Start an idea in a workspace where you’re an editor: your words, and drafts of its parts.
  • Add drafts to an idea you can edit: unknowns, things you’re not doing, a part that’s still empty.

It can’t

  • Change a word you wrote: everything it adds waits for Keep or Discard.
  • Hand an idea off, build it, or change a task.
  • Reach a workspace where an owner has turned coding agents and chat apps off.

It reads what you can read: every idea in every workspace you belong to, as a viewer or an editor, with the other pages, the UI sketches and the text of the files, and what earlier ideas ruled out or settled. It starts and adds to ideas only where you’re an editor. It never changes a word you wrote, can’t hand an idea off or build it, and costs you nothing in Walden: Walden calls no AI model to answer it.

Where the idea goes

When you ask it to start an idea, name the project if you want it in one. If you belong to more than one workspace and didn’t say which, the chat asks you where it should go, with every workspace and project in one list, and writes nothing until you’ve chosen. An idea that isn’t in a project is listed under Not in a project, like any other.

Disconnect it

Settings → Connections → Chat apps lists each app that has signed in as you, by the name it gave. Disconnect stops it at once; Allow again lets it back in. Removing the connector in the chat app itself stops it too.

An owner can turn chat apps off for a whole workspace with the switch in Settings → Coding agents, which reads Coding agents and chat apps can reach this workspace. Off, no chat app can read, start or add to an idea there. See Coding agents.

What Walden keeps

Which apps you signed in with, the name each gave when it did, and whether you disconnected one. Nothing about when you used them. What the app does with what it reads and with your conversation is between you and whoever makes it. The privacy page says it in full.