Meet Squishy: The Voice AI That Actually Builds Your Mind Map

There’s a quiet but enormous difference between an AI that can talk about your work and an AI that can do your work. Most “AI assistants” bolted into software are the first kind — a chat box that summarises, suggests, and answers questions, but ultimately hands the doing back to you. Squishy is the second kind. You talk; she takes actions on the canvas. That gap is the whole story, so let’s tell it properly.

The difference between a chatbot and an agent

A chatbot is a conversation. You ask, it answers, and any resulting work — the copying, the clicking, the building — is still yours to do. It’s a smarter help menu.

An agent is a conversation that causes things to happen. When you tell Squishy “add three branches under Marketing for paid, organic, and partnerships,” she doesn’t describe how you might do that. She does it. Three nodes appear, correctly parented, correctly coloured, while you keep talking. The doing collapses into the asking.

This is what people mean by “agentic AI,” and most products claiming it don’t actually have it. They have a chatbot wearing the word “agent” as a costume. Squishy is wired directly into the canvas’s operations — the same functions the buttons call — so anything you can do by hand, she can do by voice.

The test for a real agent is simple: after you speak, does the work exist? With Squishy, the branch is on the canvas before you’ve finished your sentence. That’s the line between talking about thinking and actually thinking faster.

What Squishy can actually do

Not a wish list — these are live capabilities:

• Create nodes and branches. “Add a child under ‘Q3 goals’ called ‘Hire two engineers.’” Done, parented correctly.
• Build entire subtrees on command. “Under ‘Launch plan,’ give me branches for pre-launch, launch day, and post-launch, with a couple of tasks each.” She constructs the whole structure.
• Move and reorganise. “Move ‘Budget’ to be a child of ‘Operations’ instead.” The node reparents and the edges redraw.
• Rename and edit. Change labels, fix wording, tidy up — all by voice.
• Summarise your structure. “What does this map look like so far?” She reads the actual structure back to you — useful when a map has grown past what you can hold in your head.
• Expand an idea. Point her at a node and ask for sub-points; she generates a first draft you then shape.
• Suggest a template. Describe your goal and she’ll set up the right starting structure.
• Argue with you. Genuinely — ask her to pressure-test a bad idea and she will. She’s been argued with. She holds up.

She knows where she is

A generic assistant answers every question from a standing start. Squishy doesn’t. She receives context about your session — which page you’re on, whether you’re logged in, how many collaborators are currently in the map — so her help is situated, not generic. Ask “who else is here?” in a shared map and she actually knows. Mention “this page” and she knows which one. The result is a partner that feels present rather than a search box that happens to talk.

Why voice, specifically

Voice isn’t a novelty here — it’s matched to the task. Mind mapping is what you do when your thoughts are arriving faster than you can organise them. In that state, the bottleneck is the interface: every time you stop to find a menu, click, and type, you lose the next thought. Voice removes that gap. You can hold a train of thought and narrate it onto the canvas at the speed you think, hands never leaving the idea. For brainstorming and brain-dumping especially, talking is simply faster than building — and Squishy turns the talking into structure in real time.

It’s also a quiet accessibility win. For anyone for whom precise mouse-and-keyboard work is tiring or difficult — motor differences, RSI, or just the end of a long day — being able to build a complex map by voice changes who the tool is for.

And she’s optional

Here’s the part that keeps the whole thing honest: Squishy is a feature, not a religion. The entire app works perfectly without ever talking to her. Click, drag, type, keyboard-shortcut your way through everything — she sits muted in the corner until you want her. We built a voice agent that can do real work and a mind mapping tool that doesn’t need it. Most people end up using her for the messy, fast, generative parts — the brain dump, the “just get it down” moments — and their hands for the careful refinement. That blend is the sweet spot.

The thing under the hood (briefly)

Without turning this into an engineering post: Squishy’s voice and conversational layer run on a real-time voice platform, and her actions are defined as a set of tools mapped directly to canvas operations. When she decides to add a node, that’s a tool call that fires the same code your click would. Her conversation persists across sessions, so she remembers the thread of what you were doing. The design goal throughout was simple — close the distance between the thought and the structure to as near zero as we can get it.

Try the part words can’t capture

You can read about an agent all day, but the moment it clicks is the first time you say something out loud and watch the map build itself in response. It’s a small piece of the future arriving early. Open a map, click the pink brain in the bottom-right, and say “help me plan my week.” See what happens.

Meet Squishy free → She’s waiting in the corner, and during beta the whole thing is free with Founder pricing locked in for life.