7 Best OpenClaw Tools for AI 3D Model Generation

We tested each skill for three things: supported output formats, ease of installation, and how well it fits into an automated agent pipeline. Generation quality matters, but agent integration matters more. A skill that produces clean STL files but requires manual downloads is less useful than one that handles the full loop from prompt to file delivery.

We also checked whether each skill works with standard 3D printing formats (STL, 3MF) and whether it can run without a local GPU. Not every team has dedicated rendering hardware, and cloud-based generation opens 3D modeling to agents running on commodity servers. The "ai 3d model generator" keyword draws roughly 8,100 monthly US searches, which tells you how many people are looking for exactly this kind of tool.

The ClawHub registry now lists over 5,400 skills across categories like Coding Agents and IDEs (1,184 skills alone), Image and Video Generation (170 skills), and more. 3D model tools are scattered across these categories, so we pulled the most relevant ones into a single comparison.

The find-stl skill by ajmwagar searches Printables for existing 3D model files and downloads them directly to your agent's workspace. Instead of generating models from scratch, it finds community-designed files that are already print-tested and ready to use.

The skill queries Printables via GraphQL, matches results against your search term, handles time-limited download URLs automatically, and writes a manifest.json with source URLs, author attribution, license details, and file hashes. That manifest is useful for compliance tracking when you need to prove where a model came from.

ClawHub Page: clawhub.ai/ajmwagar/find-stl

Strengths:

• Downloads print-ready STL, 3MF, and ZIP files with full attribution metadata
• Automates license and source tracking through manifest generation
• Works well for common requests like phone docks, enclosures, and replacement parts

Limitations:

• Limited to Printables; does not search Thingiverse or other model repositories
• Carries a "suspicious" status on ClawHub, so review the source code before installing

Best for: Agents that need to locate and download existing 3D models rather than generate new ones.

Pricing: Free.

The openclaw-openscad-3d skill by Averroeskw turns natural language descriptions into parametric 3D models using OpenSCAD. You describe a part in plain English, the LLM generates OpenSCAD code, and the skill compiles it into an STL file with a PNG preview.

This is the most format-flexible skill on the list. It reads SCAD, STL, 3MF, OBJ, OFF, AMF, DXF, and SVG as input, and exports to STL, 3MF, OFF, AMF, DXF, SVG, PNG, CSG, OBJ, PLY, and GLB. Beyond generation, it also analyzes existing STL files for physical properties like dimensions and volume, checks printability, converts between formats, and estimates printing costs. All rendering runs on CPU through OpenSCAD, so no GPU is needed.

GitHub: Averroeskw/openclaw-openscad-3d

Strengths:

• Text-to-3D via OpenSCAD code generation with PNG preview rendering
• STL analysis for dimensions, volume, and printability assessment
• Supports 12+ output formats including STL, 3MF, OBJ, and GLB

Limitations:

• Requires OpenSCAD installed locally on the host machine
• Generated models are geometric and parametric; not suitable for organic shapes or character models

Best for: Engineers and makers who need precise, parametric parts for 3D printing or mechanical prototyping.

Pricing: Free, MIT licensed.

Hitem3D by Math Magic converts reference images into textured 3D models through a structured agent workflow. Published on ClawHub in March 2026, it supports single-image, multi-view, batch processing, and portrait generation modes.

The workflow follows a clear sequence: verify API credentials, detect the task type, confirm parameters (model variant, resolution tier, output format, generation mode), submit the job via API, poll for completion, and return downloadable results with a structured summary. You can choose between geometry-only output for rapid prototyping or integrated texturing that combines geometry and surface detail in a single pass. This "structure-aware" approach produces more consistent surfaces compared to applying textures as a separate post-processing step.

ClawHub Page: clawhub.ai/lihuihui-bj/hitem3d

Strengths:

• Exports to GLB, OBJ, STL, FBX, and USDZ from a single source image
• Batch processing mode for generating multiple models in one run
• Structure-aware texture generation for consistent surface quality

Limitations:

• Requires a Hitem3D API key (pricing not publicly listed)
• Image-to-3D only; does not accept text prompts directly

Best for: Product designers and game asset creators who have reference photos and need textured 3D models.

Pricing: API key required; contact Hitem3D for pricing details.

The 3d-model-generation skill by eftalyurtseven generates 3D models from text descriptions or reference images using each::sense AI. It covers a wide range of asset types: game props, product visualizations, architectural elements, vehicles, furniture, and character models.

Output formats include GLB for web and AR/VR, GLTF for web applications, OBJ for legacy compatibility and 3D printing, FBX for animation and game engines, and USDZ for Apple AR Quick Look. The skill handles model selection and generation polling internally, so agents just send a prompt and receive a completed model file when the job finishes.

ClawHub: clawskills.sh/skills/eftalyurtseven-3d-model-generation

Strengths:

• Accepts both text prompts and reference images as input
• Five output formats covering web, AR/VR, game engines, and printing
• Handles generation polling and delivery automatically

Limitations:

• Requires an each::sense API key
• Generation quality depends heavily on prompt specificity

Best for: Teams that need quick 3D assets across multiple output formats for web, mobile, or game projects.

Pricing: Free skill; each::sense API usage costs apply.

Store and share generated 3D models in one workspace

Free 50GB storage for your 3D assets. No credit card, no expiration. Connect any OpenClaw skill to a shared workspace with built-in semantic search.

Start 14-Day Trial

The cad-agent skill by clawd-maf is a Dockerized rendering server that lets agents write build123d modeling code and see what they're building. The agent sends parametric modeling commands over HTTP, the server executes them inside a container, and VTK renders the result as a PNG image. The agent inspects the render, adjusts the code, and iterates until the part looks right.

This visual feedback loop is what sets cad-agent apart from the other generators on this list. Instead of producing a model blindly and hoping the output matches the intent, agents can check each iteration and tweak dimensions, features, or tolerances before committing to a final export. The trade-off is speed: each render cycle takes a round trip, so complex parts require patience.

ClawHub: clawskills.sh/skills/clawd-maf-cad-agent

Strengths:

• Visual feedback loop lets agents inspect rendered previews at each design step
• Uses build123d for precise, parametric CAD rather than mesh-based approximation
• Runs in Docker, keeping the host machine clean

Limitations:

• Requires Docker installed and running
• Agents need some familiarity with build123d's API to produce useful output

Best for: Mechanical engineers and hardware teams building precise, parametric parts through iterative visual refinement.

Pricing: Free, open source.

3D AI Studio provides a free, MIT-licensed OpenClaw skill backed by three different AI models: TRELLIS.2, Hunyuan Rapid, and Hunyuan Pro. Each model trades off speed against quality. Hunyuan Rapid generates models in roughly 25 seconds, while Hunyuan Pro takes up to 6 minutes for higher-detail output.

The skill accepts both text prompts and reference images. Primary output is GLB with optional PBR textures, which can be converted to STL for 3D printing, OBJ, or FBX for game engines. Conversion runs either locally through Blender or via the 3D AI Studio conversion API. All generation happens in the cloud, so no local GPU is needed. The multi-model approach is useful when you're prototyping: use the fast model for quick iterations, then switch to the high-quality model for final assets.

Source: 3daistudio.com/Models/OpenClaw-3D

Strengths:

• Three model options let you balance speed and detail per task
• PBR texture support for production-ready assets
• Cloud-based processing with no GPU requirement

Limitations:

• Requires a 3D AI Studio API key
• GLB-to-STL conversion needs either Blender or the conversion API

Best for: Teams that want model flexibility and prefer not to lock into a single generation backend.

Pricing: Free skill (MIT); API usage costs may apply.

Fastio is not a 3D model generator. It's the workspace where generated models land after the pipeline finishes. When agents produce STL, GLB, or OBJ files through any of the skills above, those files need persistent storage, organization, and a reliable way to hand them off to human teammates.

Fastio's MCP server exposes 19 tools for file management, uploads, AI chat, and workspace operations. Intelligence Mode auto-indexes uploaded assets for semantic search, so finding "the bracket model with 4mm wall thickness" works without browsing through folders. Agents create workspaces, organize generated files by project or client, and transfer ownership to human leads when the project is ready for review.

The Business Trial includes 50GB storage, included credits, and 5 workspaces with no credit card required. For teams running generation pipelines that produce dozens of models per session, the persistent workspace means agents pick up where they left off without re-downloading or re-generating previous assets.

ClawHub Page: clawhub.ai/dbalve/fast-io

Strengths:

• generous storage and monthly credits during the trial with no credit card required to start
• Semantic search across indexed model files via Intelligence Mode
• Ownership transfer: agents build the asset library, humans receive it

Limitations:

• large max file size on the free tier
• Intelligence Mode consumes AI credits when indexing large file sets

Best for: Teams connecting agent-generated 3D assets to human review and production workflows.

Pricing: Business Trial. Pro plans at fast.io/pricing.

If you already have models somewhere and just need to find them, start with find-stl. It searches Printables and delivers print-ready files with full attribution tracking. No generation, no API keys, no waiting.

For generating new models from text, the choice depends on what you're building. openclaw-openscad-3d handles precise parametric parts (brackets, enclosures, mounting plates) where exact dimensions matter. 3d-model-generation and 3D AI Studio handle organic shapes, characters, and artistic assets better. Hitem3D is the pick when you have reference photos and need to recreate them in 3D.

For iterative engineering work where the agent needs to see what it's building and adjust on the fly, cad-agent gives visual feedback at each design step. It's the closest thing to an engineer working at a CAD workstation.

Once your agents generate files, add Fastio to persist, search, and share those assets with your team. The free tier covers most prototyping workflows with 50GB of storage and built-in semantic search across indexed files.

All seven tools are free to install. The ones that call external generation APIs (Hitem3D, 3d-model-generation, 3D AI Studio) need API keys with associated usage costs. find-stl, openclaw-openscad-3d, cad-agent, and Fastio run without paid API dependencies.