The First Card Is the Hardest

The blank canvas is harder than it looks.

You know what you’re building. You’ve been thinking about it for days. But opening a new canvas and placing the first card is surprisingly slow. Where does it go? What are the main components? How do they connect? You have the idea — you just don’t have the structure yet.

This is the problem Stoker is designed to solve.

The structure problem

Most AI features in planning tools generate text. They write acceptance criteria, summarise requirements, expand a bullet point into a paragraph. This is useful, but it misses something.

The hard work at the start of planning isn’t writing — it’s decomposition. Breaking an idea into its parts. Figuring out what the components are, how they depend on each other, and what order they should come in. That’s a structural question, not a prose question, and text generation doesn’t answer it.

Stoker generates a canvas. You describe your project in a sentence — “Plan a user onboarding flow with email verification and profile setup” — and you get back a set of nodes connected by edges. An actual structure you can react to, not a paragraph you have to read and then translate into one.

Reacting is faster than creating

There’s a well-known effect in writing: editing is faster than drafting. A blank page is the hardest starting point; a page with words — even wrong words — is easier to work with. You read it, disagree with parts, cross things out, add what’s missing. The thinking accelerates.

The same is true for canvas planning. A canvas with ten nodes — even if some of them are wrong — is a better starting point than an empty one. You can see the structure and immediately form opinions. This step is missing. That dependency goes the other way. These two things are actually the same component. Those reactions are the planning process. Stoker just starts it faster.

What Stoker doesn’t do

Stoker doesn’t plan for you. The generated canvas is a scaffold, not a result. You’ll rename nodes, delete ones that don’t apply, add ones that are missing, and redraw edges to reflect how your project actually works.

That’s by design. The goal isn’t to produce a correct plan automatically — it’s to get past the blank canvas so you can do the actual thinking.

Always additive

Stoker always adds to your canvas, never replaces it. You can run it on a project you’ve already started and it will place its output alongside your existing work. Before anything lands, you see a preview: a mini canvas showing the nodes and edges about to be created. Click Accept when you’re ready.

The Improve button lets you refine the result before committing. Tell Stoker what’s missing or what to change, and regenerate with the previous output as context. You can iterate as many times as you like before anything touches the canvas.

Bring your own key

Stoker uses your own OpenAI or Anthropic API key — stored server-side, never in the browser. You choose the model. Your data goes to your AI provider under your account.

Add your key in Settings → AI Integration (or Org Settings → AI Integration if you’re setting it up for a team). From any canvas, click the ✦ Stoker button to open the panel and start generating.

The first card is the hardest. After that, the planning takes care of itself.