Work with AI in more than words.

Whiteboard, design studio, code IDE, notebooks, sheets, workflows and chat — all on one project, with AI that can see what you’re working on — it knows what you mean: instruct it visually, not just in text, and collaborate on documents, designs and diagrams.

Free versions available. No card.

Whiteboard. Sketch it, or let AI draw it — handwriting becomes editable text, and every shape is a real object you can restyle, rearrange, or run.

Your AI can’t see what you’re working on.

So you screenshot the design. Describe the diagram in words. Re-explain the layout it can’t see. Then open the next tool and do it again.

You shouldn’t have to describe your own work to a machine that could just look at it.

In Circuitry it can. The agent reads the document you’re in — the actual shapes, layers, nodes and cells, and which one you’ve selected — and can look at how it renders when that helps. It’s reading the structure, not just a picture of it. So you point at what you mean instead of describing it, and carry on talking.

One interface can’t fit every kind of thought.

So Circuitry gives every kind of thinking its own surface.

Not nine apps — one project, seen nine ways. Chat included — and the coding assistant you already use runs right here, seeing the document you’re working in.

Circuitry Studio desktop app — free/basic mode.

The right editor for each job.

  • Draw it on the whiteboard.
  • Build it in the IDE.
  • Explain it in a notebook.
  • Automate it in a workflow.

It’s still the same project.

No re-explaining.

No rebuilding context.

No starting over.

You show AI. AI shows you.

YOU SHOW AI

Wire it up. Then say one line.

The inputs carry the context, so the instruction can be a sentence instead of an essay. Select what you mean and say what you want.

AI SHOWS YOU

Ask what your payment flow does. It draws it.

The agent reads the code and draws it back to you — and flags what worries it, in red. Edit the drawing. It changes the code.

Every idea gets the surface it deserves.

Draw it instead of describing it.

Sketch a system on the board. It becomes a diagram you can arrange — then a workflow that runs.

Point instead of explaining.

Circle it, select it, highlight it. Then say what you want. What you select is what your agent sees — nothing to re-describe, and nothing you didn’t choose.

Write it by hand.

Handwriting becomes editable text as you write. A rough drawing comes back as a finished image.

Work the way the task wants.

Some problems are spatial. Some are numerical. Some are code. Use the surface that fits — not the one you’re stuck with.

Keep the final say.

Every edit the agent makes arrives as an accept-or-reject diff. Nothing changes behind your back.

Point it at your own files.

Loop a folder on your own disk through AI — background removal, conversion, resizing — and write the results straight back. No uploading, no per-task cloud bill.

No exporting.

No importing.

No lost context.

From a rough sketch to a running API — without leaving.

HandwritingText
SketchDesign
SketchWorkflow
DesignCode
WorkflowLive API

Every one of those steps has a proper tool. Every one of those tools has intelligence in it. And all of them work on the same project.

  1. 01Every step in a workflow picks its own language.Drop in a code node and choose JavaScript, Python, Three.js, Shell, Go, Rust, C, C++ or Zig. Changed from a dropdown, not a rebuild — and steps in different languages sit in the same flow.
  2. 02Ask the node for what you want.Every node has an AI wizard. Type “Mandelbrot fractal” into a Python code node and it writes the code — in the node, in the language you already chose.
  3. 03Run it. See the output.Connect a code node to an image node and its output renders on the canvas. Real Python here — numpy and PIL computing Mandelbrot and Julia sets — with each result feeding the next step.
  4. 04Ship it as a live endpoint.Compiled Rust behind a webhook, run and green. The graph you drew is the thing answering the request — no separate deploy step, no rewrite.

Compiled languages (Go, Rust, C, C++, Zig) run on paid plans.

One project. Many specialised surfaces. Your coding agent sees the one you’re in.

Start free. Add the AI when you want it.

The free plan isn’t a trial. It’s the visual workspace in the web, and a full code IDE on your own machine, for as long as you like. Open either and see what it is.

Free, in the web

The whiteboard, design studio, workflows, sheets and notebooks, with Python and JavaScript steps. No install, no card — just a limit on saved files.

Free, on your own machine

Download the app and get a full code IDE: your own projects and files, real terminals, and your own command-line coding agent working right there in it.

Add AI from $9

Cloud models, agents and wizards, plus automations you can deploy so they run in the cloud on a schedule or whenever another app calls them.

The whole studio from $20

An AI co-pilot in every editor, your designs and workflows saved to your own disk, your other machines connected, and working alongside someone else in the same document. iPad included in the same subscription.

Leave whenever

Your work exports as real code in the language you actually use. Nothing is held hostage.

Cloud models, on-device models and command-line coding agents — pick per conversation. Local models and command-line agents run in the desktop app, which is a free download.

And it’s coming to iPad. Not a companion app — the whole studio, pencil-first.

One project. Many specialised surfaces. Your coding agent sees the one you’re in.

Think with AI. Not just talk to it.

Free on your machine and free in the web. No card either way.