What You Can Do With the MCP
Every tool in the 3D AI Studio MCP connector: 3D generation models, image generation and editing, low poly, UVs, format conversion and PBR materials, with costs.
This page lists everything your assistant can do once it is connected to 3D AI Studio, with the credit cost of each action and prompts you can copy.
You never have to call tools by name. Just describe what you want, and the assistant picks the right tool. The tool names are listed so you can recognize them in your assistant's activity log, or ask for a specific one.
Costs on this page are the standard prices at the time of writing. Your assistant always has the current prices: ask "which 3D AI Studio models are available and what do they cost?" and it will check for you. Every action costs the same credits as on the website.
At a glance
| Category | What you can ask for |
|---|---|
| 3D generation | 3D models from text, one image, or several views |
| Images | Generate images from text, edit images with instructions |
| 3D processing | Low poly versions, UV unwrapping, format conversion and scaling |
| Materials | Tileable PBR materials from text or from a photo |
| Image utilities | Background removal |
| Library and projects | Find past generations, show results, organize into projects |
| Uploads | Hand your own photos and sketches to the assistant |
| Account | Credit balance and the model catalog |
3D generation
All 3D models support text to 3D (a description), image to 3D (one image), and multi view to 3D (several images of the same object from different sides). A 3D model usually takes 1 to 5 minutes. Your assistant checks on it until it is done.
| Model | Best for | Inputs | Typical cost |
|---|---|---|---|
Prism 3.1 (recommended)generate_prism_3_1 | Best overall quality. The default choice for almost everything | Text, 1 image, or 2 to 4 views | 35 credits |
P2generate_p2 | Low poly, game ready topology, with optional native quad mesh output | Text, 1 image, or 2 to 4 views | 70 credits with texture |
P1generate_p1 | Previous generation low poly model, cheaper than P2 | Text, 1 image, or 2 to 4 views | 60 credits with texture |
Hunyuan 3.1generate_hunyuan_3_1 | Clean, even meshes at a flat price. PBR included | Text, 1 image, or up to 8 views | 40 credits flat |
Meshy 7.1generate_meshy_7 | Very high resolution 4K and 8K textures and ultra detailed geometry | 1 image, or 2 to 4 views (no text) | 40 textured, 35 mesh only |
Options and surcharges
You can ask for these in plain language ("with extreme texture detail", "as a quad mesh", "without texture"):
| Option | Models | Extra cost |
|---|---|---|
| Detailed texture / extreme texture | Prism 3.1 (+5 / +10), P2 and P1 (+10 / +15) | As listed |
| Detailed geometry | Prism 3.1 | +10 |
| Multi view input | Prism 3.1, P2, P1, Meshy 7.1, Hunyuan 3.1 | +5 |
| No texture (geometry only) | P2 and P1 (saves the 20 credit texture charge), Meshy 7.1 (35 instead of 40) | Cheaper |
| Quad mesh output | P2 | Free |
| Newest texture model (experimental), with a fast texture option and lighting removal for photos | Prism 3.1 and P2 | Free |
| 4K / 8K textures | Meshy 7.1 | +10 / +15 |
| Ultra geometry | Meshy 7.1 | +10 (2K) / +15 (4K, single image) |
| Low poly mode | Meshy 7.1 | 20 textured / 15 mesh only, plus texture surcharges |
| Low poly, geometry only and sketch modes | Hunyuan 3.1 | Included in the flat price |
For multi view input, order matters. Give the views in the order front, left, back, right. One good image gives great results in most cases, so you only need multiple views when the back or sides of your object are very different from the front.
Example prompts
- "Make a low poly fox for my game."
- "Turn this product photo into a 3D model with detailed textures."
- "Create a three headed wolf statue I can 3D print."
- "Generate a sci-fi crate with P2 as a quad mesh."
- "Use Meshy with 8K textures on this image of a leather boot."
Images
Images take seconds and are a great first step: design the look as an image, then turn it into 3D.
| Model | What it does | Cost per image |
|---|---|---|
Nano Banana 2.1 (recommended for generating)generate_image_nano_2_1edit_image_nano_2_1 | Fast, consistent images. Also edits with up to 14 input images | 6 (1K), 9 (2K), 13 (4K). Edits: +1 per started group of 3 input images. +3 with Google search grounding |
GPT Image 2 (recommended for editing)generate_image_gptedit_image_gpt | Very precise instruction following and text rendering. Edits up to 10 input images, optionally with a mask | Low quality 3 / 5 / 8, medium 6 / 14 / 25 (1K / 2K / 4K), high 30 |
Nano Banana 2 Litegenerate_image_nano_lite | The cheapest image generation | 5 flat (+2 with Google search grounding) |
Nano Banana 2edit_image_nano | Previous generation editor | 7 (0.5K), 10 (1K), 14 (2K), 20 (4K). +2 with Google search grounding |
You can generate 1 to 4 images per request. Generated images can be used directly as input for 3D generation.
Example prompts
- "Generate a cozy wizard hut on a white background, then turn it into a 3D model."
- "Edit this photo: remove the person in the background."
- "Make three variations of this character in different outfits."
- "Generate a front view of a cartoon robot so I can turn it into 3D."
3D processing
These work on a model you already have, either one you just generated or one from your library. Each produces a new model, so the original stays untouched.
| Action | What it does | Cost |
|---|---|---|
Low poly versionlow_poly_model | Clean, lighter retopology for games, AR, VR and real time use. The default version (3.0) can unwrap UVs in the same step for free. Takes 1 to 3 minutes | 30 (35 as quad), 25 for version 2.0, 15 for version 1.0 (basic decimation) |
UV unwrapunwrap_uvs | Clean UV layout while keeping textures, rig and animation. For models up to 80,000 triangles (make a low poly version first for denser models). Also saves a free preview of the UV layout | 15 flat |
Convert formatconvert_model | GLB, OBJ, OBJ with textures (zip), FBX, FBX with textures (zip), STL, PLY, USDZ, 3MF. Can scale the model down to a maximum size in cm or inches, for 3D printing | 1 |
Example prompts
- "Make a low poly version of that model for my mobile game."
- "Unwrap the UVs of my last model."
- "Convert it to STL and scale it to 12 cm tall."
- "Export the model as FBX with textures."
Materials
| Action | Output | Cost |
|---|---|---|
Material from textgenerate_material | A seamlessly tiling PBR material: base texture plus base color, normal, roughness, metalness and height maps | 10 for the whole set |
Material from a photomaterial_from_image | The same PBR maps extracted from a photo or texture image. Works best with flat, evenly lit close ups of a surface | 7 for the whole set |
Materials are for surfaces (brick, rusted metal, moss, fabric) to use in a game engine or 3D app. They do not texture an existing 3D model.
Example prompts
- "Create a mossy cobblestone material for my level."
- "Turn this photo of oak bark into a PBR material."
Image utilities
| Action | What it does | Cost |
|---|---|---|
Remove backgroundremove_background | Keeps the subject on a transparent background. The cheapest option works for most images; the most expensive one is best for product photos | 1 to 3, depending on the quality option |
You do not need to remove the background before turning an image into 3D. The 3D models handle backgrounds on their own.
Library and projects
| Action | Tool | Cost |
|---|---|---|
| Browse and search your generations | list_generations | Free |
| Check a generation's progress or result | get_generation | Free |
| Show a finished result in the chat | show_generation | Free |
| List your dashboard projects | list_projects | Free |
| Create a project | create_project | Free |
| Move a generation into a project | assign_to_project | Free |
Everything your assistant creates lands in your dashboard. You can also work with models you made on the website: "find my dragon model from last week and convert it to FBX".
Example prompts
- "Show me my last five 3D models."
- "Create a project called Forest Level and put all models from today in it."
- "Generate a barrel, a crate and a lantern, and put them in my Tavern project."
Uploads
| Action | Tool | Cost |
|---|---|---|
| Hand an image from your device to 3D AI Studio | upload_image | Free |
| List your recent uploads | list_recent_uploads | Free |
Your assistant cannot pass images from the chat to 3D AI Studio directly. Instead, it opens an upload card in the chat (or gives you an upload link). See Images & Uploads for how it works.
Account
| Action | Tool | Cost |
|---|---|---|
| Check your credit balance | get_credit_balance | Free |
| List models with their inputs and costs | list_available_models | Free |
Example workflows
Game asset, from idea to engine
"Generate a front view image of a stylized treasure chest, turn it into a 3D model with P2 as a quad mesh, then convert it to FBX with textures."
3D print
"Turn this sketch into a 3D model, then convert it to STL scaled to 8 cm."
Product visualization
"Make a 3D model of this sneaker photo with detailed textures and show it to me."
Level dressing
"Create a project called Swamp, generate a mossy log, a lantern and a wooden sign, put them all in it, and make a mossy stone material for the ground."
What is only on the website
The connector focuses on generation and the most common follow up steps. For these, open the model in your dashboard:
- Advanced generation settings and all other 3D models on the platform
- Texturing, the Texture Painter, rigging and animation, segmentation and remeshing with all options
- 3D printing tools such as Mesh Repair, Add Base and Export for 3D Printing
- Flow, the node based editor
See the rest of the documentation for those tools.
Frequently Asked Questions
If you are unsure, do not specify one. Your assistant uses Prism 3.1 by default, which gives the best overall quality. Ask for P2 for low poly or quad meshes, Hunyuan 3.1 for clean meshes at a flat price, and Meshy 7.1 for very high resolution textures from an image.
Yes. Your assistant can browse your library and use any of your models or images as input, for example to convert, unwrap or make a low poly version.
The connector asks assistants not to talk about credits unless you ask, to keep conversations focused. Ask "how much will this cost?" or "what's my balance?" at any time and it will tell you.
Other MCP Clients
How to add 3D AI Studio to any MCP client with remote server support: the server URL, transport, OAuth sign-in, and a bridge for clients that only run local servers.
Images & Uploads
How to give your AI assistant a photo, sketch or reference image for 3D generation or editing: the inline upload card, the upload page, links and library images.