Product

# Hydrilla Agent, the AI 3D agent for Blender

Hydrilla Agent is the AI 3D agent that works in your own Blender. Tell it what to build or change, and it does the work in your open scene, or makes a real 3D model with BlueFox from a prompt or an image.

Talk to it the way you would talk to a 3D artist: what you want, how it should look, and where it goes. It looks at your scene, does the work, checks the result, and shows you each step as it goes.

## What it can do

Hydrilla Agent connects to Blender through the Hydrilla add-on, so it works on the scene you already have open.

- **Make 3D models.** Describe an object and it makes a detailed model with BlueFox, then puts it straight into your Blender scene.
- **Work from images and files.** Attach a photo or concept art and it turns the image into a 3D model, or add a PDF or text file, such as a brief or a spec, for it to follow.
- **Build and change your scene.** Ask for a table with four chairs, warmer lights, or a camera on the hero prop. It builds and edits in Blender for you.
- **Understand your scene.** Ask what is in the scene, how big something is, or what changed. It reads the scene and looks at the viewport.

## How a session works

1. Open Blender with the Hydrilla add-on and paste your Hydrilla API key. It connects on its own.
2. Tell the agent what you want, and attach any images, PDFs or text files it should work from.
3. For a new object, choose how to make it: a detailed model with Hydrilla AI, or a quick build from simple shapes in Blender.
4. Approve the model and its cost, or turn on Auto approve to skip that step. It generates, then lands in your scene, and you keep refining in the same chat.

## You stay in control

With Always ask, the default, nothing that costs credits starts without your approval: you see the prompt, the models, the quality and the cost first. Switch to Auto approve under the message box when you want models to start right away. Every step shows in the chat as it happens, and Ctrl+Z in Blender undoes any change the agent made. Each chat is saved with a short name, so you can find it in your history and carry on where you left off.

## Where Hydrilla Agent works

Hydrilla Agent runs in the Hydrilla app and works in Blender through the Hydrilla add-on (version 2.3 or newer). [Set up the Blender add-on →](https://hydrilla.ai/integrations/blender)

## Questions about Hydrilla Agent

What does Hydrilla Agent cost?
Chatting runs on the AI model you pick with your own API key (Anthropic, OpenAI, Google or OpenRouter), billed by that provider. 3D models made with BlueFox use Hydrilla credits, and the agent asks before it spends any unless you turn on Auto approve.

Which AI models can it use?
Any chat model from Anthropic, OpenAI, Google or OpenRouter that you have a key for. You pick the model under the message box; models marked with the eye icon can also see images and read PDFs.

Does it work without Blender?
Yes. You can still chat and make 3D models with BlueFox, and import them into Blender later. Building and editing a scene needs Blender with the Hydrilla add-on.

What files can I attach?
Up to 4 per message: images (PNG, JPG or WebP), PDFs up to 3 MB, and text files up to 200 KB, such as notes, CSV, JSON or scripts. You can also paste or drop them into the chat. Blender and 3D model files cannot be attached; open them in Blender and the agent works on the scene.

When should I use text to 3D instead?
When you just want one model and do not need a scene. [Text to 3D →](https://hydrilla.ai/text-to-3d) starts from a prompt, and [image to 3D →](https://hydrilla.ai/image-to-3d) starts from a reference.

How is Hydrilla Agent different from the MCP server?
Hydrilla Agent is the agent inside Hydrilla. The MCP server lets your own AI agent, such as Claude Code or Cursor, drive Hydrilla from outside. [MCP server →](https://hydrilla.ai/integrations/mcp)

## Build your next scene

Open Hydrilla Agent and tell it what to make.
