Quolla

A canvas for thinking with AI

Where the first thought lands.

On the left, a linear AI chat. On the right, a canvas. Drop the lines worth keeping onto the map, add your own notes and questions, then connect them — the connected nodes become the memory the AI reasons from. Bring your own key; everything stays on your machine.

macOS & Windows · or run it in your browser

Quolla · Map 1
OPENAI · GPT-5.2

How should I structure the essay?

Start from your central claim and build outward. One thing to hold onto: the act of placing it is itself a memory hook+ Add to map — what you set down by hand is what stays.

Select text → Add to map

the problem

A thread only moves one way.

Ask, answer, scroll. In a linear chat, every fork in your reasoning flattens into the same column — and the structure disappears the moment you make it.

Linear chatscrolls away
How should I structure the essay?Start with your central claim, then build outward from it.What about counterarguments?Address the strongest one early so the reader trusts you.Which sources back that up?The 2019 longitudinal study is your best anchor.Wait — go back to the opening idea.It has scrolled up. Let me restate it for you…How should I structure the essay?Start with your central claim, then build outward from it.What about counterarguments?Address the strongest one early so the reader trusts you.Which sources back that up?The 2019 longitudinal study is your best anchor.Wait — go back to the opening idea.It has scrolled up. Let me restate it for you…
The same thinking, mappedstays put
Chat
EXCERPTA memory hook
NOTEHand moved, it sticks
QUESTIONWhy do I forget?

Quolla keeps the pieces you choose where you can see them.

how it works

From a line of chat to a memory you control.

Chat as usual, keep the pieces worth holding onto, and connect them into the context the AI works from.

  1. 01

    Think it through in chat

    A normal linear AI conversation on the left. Ask, get answers, follow the thread — your key, your provider.

  2. 02

    Keep what matters on the map

    Select any line and add it to the canvas as a node, or place your own note or question. Nothing lands on the map unless you put it there.

  3. 03

    Connect nodes into memory

    The nodes you connect become the memory the AI reasons from. Reconnect them — or scope to the map — to reshape its context.

the control surface

You shape the memory by moving the map.

The map is the control surface, not a picture. Link a node to the origin and it enters the model's context; unlink it and it's gone. Regroup or reconnect nodes and you reshape what the AI reasons from. You edit memory by moving things on the canvas, not by managing a settings panel.

map.canvas
originChat

Tap a node to add or remove it · Drag from a node's edge to connect

What the model sees

  • AI EXCERPTThe act of placing it is itself a memory hook.
  • NOTELike writing on paper: it sticks because your hand moved.

The AI reasons from 2 nodes you kept on the map — nothing else.

Scope the memory to a map.

Turn on SOURCES: MAP ONLY and the AI sees only what lives on the map — the excerpts, notes, and questions you kept. It stops drawing on the whole conversation and reasons from the context you built by hand.

Every node remembers where it came from.

Each node keeps an origin link back to the message that produced it. Follow ↗ origin to see why a thought is on the map; nothing detaches from its source.

Your data stays plain and portable.

Export any map — chats, nodes, and links — as plain JSON, and bring it back anytime. Quolla never holds your thinking hostage.

Privacy & BYOK

Your thinking never leaves your machine.

Quolla is built so the private part stays private. The key is yours, the data is yours, and nothing phones home.

  • Your key, your account

    Quolla is bring-your-own-key. Your API key talks directly to the AI provider you choose. You hold the relationship and the billing.

  • Your data stays local

    Maps live on your machine. Nothing routes through a Quolla server, because there isn't one in the loop.

  • No telemetry

    No usage tracking, no analytics calls watching what you write. The app does its job and stays quiet.

pricing

Start free. Go further when you outgrow one map.

Free is the whole instrument on a single map. You only hit a wall when you want a second one — which means Quolla is already working for you.

Free

The whole instrument, one map.

Keep chat excerpts and your own notes on the map, connect them into the memory the AI reasons from, and restructure as much as you want. Origin tracing and JSON export are included. Bring your own key; everything stays on your machine.

  • Drop chat excerpts onto the map, or add your own notes and questions
  • Connect nodes into the memory the AI reasons from
  • Origin tracing on every node
  • Plain-JSON export and import

Pro

One purchase

One purchase, no subscription.

  • Unlimited maps. Keep a separate space for every line of thinking.
  • Scope the memory. Turn on SOURCES: MAP ONLY so the AI reasons strictly from the nodes on the map, nothing else.
  • Drop in your own documents. Add PDFs and notes to a map and the AI reasons grounded in them — a private knowledge base over your own sources, still local, still your key.

Price coming soon · one purchase, no subscription

Buy Quolla Pro

Later— optional cloud sync. A subscription to sync your maps across devices is on the way. Entirely opt-in: the local, no-server Quolla you start with never changes.

Open a map in your browser.

Runs in your browser. Bring your own key. Nothing to install — start from one node and see where it goes.

Download

Get Quolla for your desktop.

Free to download. You'll add your own API key on first run — that's what powers the thinking.

All versions & release notes

First-time Windows users may see a security warning — it's safe. Here's why and how to proceed →

You bring the key
Quolla needs an API key from your AI provider. Paste it on first run; it's stored locally and used only to talk to that provider.
First-launch warning (for now)
macOS is signed and notarized. Windows may show a first-run warning while code signing is on the roadmap; see the installation notes for the exact steps.
System requirements
macOS 12 or later · Windows 10/11 (64-bit).