Introducing Planaria: Text-to-Component AI for Grasshopper

[planaria]

Hi everyone,
I’m excited to introduce Planaria — a tool I’ve been building over the past few months.
Quick Overview:
No API Key required — We host the AI models
No coding knowledge needed — Describe your logic in natural language
Safe by design — Generates isolated components; never touches your model
Windows only (Rhino 7 & 8)
Google Account required for login

What is Planaria?
A Text-to-Component AI system. You describe what you need, and it generates a compiled Grasshopper component based on your description.

Why “Planaria”? The Planarian flatworm is famous for its incredible regeneration abilities. This symbolizes the core of our system: Auto-Debugging. When generated code has errors, the system attempts to “regenerate” and fix itself, much like its namesake, to deliver a working component.

Philosophy: Tool Maker, Not Operator
As a parametric designer working in the industry, I follow a strict principle: I never let an LLM directly control my Rhino model.

While AI Agents (MCP) are powerful, the hallucination risk is too high for precise architectural projects. I can’t risk an AI silently moving a wall or deleting geometry without my knowledge.
Planaria is designed for safety: Instead of letting AI be the “Operator,” we let it be the “Tool Maker”. It generates isolated components you can inspect and test on the canvas. If it works, use it. If not, delete it. Your model stays untouched.

Core Features

1. Natural Language + Auto-Debugging
   Type your request, wait 10–60 seconds (depending on complexity), and the system self-corrects errors to deliver a valid component.
   Tip: Describe geometric logic step-by-step for best results:
   ✓ “Divide the surface into a grid, then place boxes at center points”
   ✗ “Make it look parametric”

2. The Lab (Community)
   The greatness of Grasshopper lies in its community. Planaria Lab lets you share your components publicly, or stand on the shoulders of giants — download others’ work and modify it via prompts to fit your needs.

   

How to Install:
Open Rhino 7 or 8 (Windows)
Type PackageManager
Search for Planaria
Install and Restart Rhino

Website : https://planaria.ai

Food4Rhino : https://www.food4rhino.com/en/app/planaria

Beta Notice & Call for Feedback
Planaria is in early public testing — you may encounter some rough edges. I released it early because I believe in this community’s power to shape tools. Whether it’s a bug report or a feature request, I’d love to hear from you.

Specifically, I’m looking to optimize the AI’s training: Are there specific algorithm libraries or geometry logic you want improved? Let me know, and I’ll prioritize them in the next update.

Cheers,

Hsuan-Ku Lin
Architect & Creator of Planaria

[planaria]

V1.0.81 Update: Your credits now go much further!

We realized that you shouldn’t be charged when the AI is fixing its own compilation errors. So we changed the backend logic.

From now on, all “Auto-Fix” operations are on the house.

This means your usage count now strictly reflects the number of components you create, not the number of attempts the AI makes. This effectively multiplies your testing capacity!

Give it a try and let me know if it helps your workflow.

[GHwiz]

I hvnt tested it but i like the concept. I know there is already similar tools but this one looks really nice, this is what the future of AI and grasshopper looks like.

”the system self-corrects errors to deliver a valid component.” this part is key, i find myself doing this always with chatgpt when coding a component.

what would be cool in the future is if the component could break out into creating the cluster of components on screen. Like some sort of “unpackage”. and then can “package it back” after some manual edits. Also, in my opinion a time-based subscription might work better than paying by component created.

Just some ideas, really good stuff!

[planaria]

@GHwiz Thanks for the feedback!

Just to clarify the workflow: Planaria compiles the logic into a native C# component. While it doesn’t ‘explode’ into standard Grasshopper nodes, it is designed to be extremely easy to customize simply by updating your text prompt.

The real secret sauce is in the fine-tuning. We specifically trained the model on RhinoCommon and complex geometric algorithms. On top of that, we’ve implemented optimized encapsulation for various Grasshopper Hint Types, ensuring the code handles data inputs and type casting much more reliably than generic AI.

As an industry professional myself, I built this because I needed it—and I’m honestly amazed by how much time it has already saved me.

Regarding pricing: We do offer a subscription model! The Free Plan currently includes 30 generations per month, which should be plenty for you to give it a solid test run.

Let me know how it goes!

[crz_06]

looks interesting and nicely designed. Nice unique take on the safety aspect and the addition of the eco-system is a nice touch. I will test it over the Christmas break.

• So without seeing or having used it some question. Can a user see the formatted generated C# code or an up to date prompt that describes the nodes actual function, i.e. how you or other user can best to edit /tweak it in the future.
• You mentioned it’s trained on C#, what happens if i provide full block of python code and ask it to convert as to getter better performance

[planaria]

Hi @crz_06,

Regarding your questions:

1. Visibility & Documentation: The component actually generates its own documentation. The AI automatically configures Grasshopper Type Hints and generates custom tooltips for every input based on the logic it wrote.

As you can see in the screenshot, even without reading the raw C#, you can hover over an input (like offsetFactor) and see exactly what it does and what value range is expected. This “UI-as-Documentation” gives you the immediate context needed to tweak your prompt effectively.

2. Python to C# Conversion: Yes. This is actually a primary use case. Since the model understands the underlying RhinoCommon API, it translates logic between languages very effectively. You can paste your full Python block into the prompt and ask it to “Rewrite this in optimized C#.”

Let me know if you run into any issues during your tests. Merry Christmas!

[Cumberland]

Please put this components into the “Maths“ panel (the second panel from the left). I already have plenty of component tabs that don’t fit on my screen majority with only 2-5 components in them. A lot of wasted space. I do need to stretch the Grasshopper on the length of two monitors to be able to select a components tab.

[planaria]

Really appreciate you taking the time to share this suggestion!

I’ve thought about it, and decided to keep Planaria in its own tab for now — mainly because new users often discover plugins through the component tabs, and having a dedicated space makes it easier to find.

That said, I’ve also added a Planaria menu in the top menu bar and a toggleable floating widget on the canvas, so you can access it quickly without going through the tab.

Hope that helps! Let me know if you have other ideas.