M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
N

Nova3D

Community
Code-native 3D generation — assets are executable programs, not opaque meshes.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #295 of 438
Stars
★ 729 Very popular
Transport
stdio (local process)
Runtime
Python · Hosted (no local runtime)
Credentials
Optional API key
License
MIT
Last commit
46FMRS · D

Nova3D MCP server offers a unique approach to 3D generation, producing assets that are programmatic and part-aware, distinguishing it from traditional mesh generators. It is best suited for developers and AI agents requiring editable, animatable structured 3D models. The server uses a hosted service, so an API key is required. Users should consider ongoing availability and costs due to the external service dependency. The implementation is interesting for technically savvy users with an interest in Blender and generative AI.

Strongest · Maintenance 11/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
10/20
Maintenance
11/20
Documentation
8/20
Setup experience
9/20
Why each score
Reliability 8/20
Evidence is limited to the server manifest and README. The manifest is internally consistent (0.4.0, PyPI, uvx, stdio), but no tool list, server source, CI, or tests were supplied, so the init/tool-list handshake and whether declared behavior matches reality cannot be verified. Under static calibration the happy path is only plausible, and testing/error-handling evidence is missing; score 8.
Security and permissions 10/20
NOVA3D_TOKEN is declared optional and secret, with browser sign-in preferred; no red-line issues such as malware, credential theft, or covert exfiltration are evidenced. However, data-flow disclosure for prompts/images sent to the hosted backend, a security policy, and confirmation flows for sensitive operations are absent, and least-privilege evidence is thin; score 10.
Maintenance 11/20
MIT license is explicit, the repo is not archived, only 1 open issue, and manifest version 0.4.0 shows versioning practice. But no commit history, release cadence, dependency updates, or security-response channel appears in the evidence, leaving governance and update path incomplete; score 11.
Documentation 8/20
The main README richly covers product vision, comparisons, and media, but install, auth, tool parameters, limits, cost, and troubleshooting for the MCP server live in the referenced mcp/README.md, which was not supplied. Only the manifest's env-var descriptions are available, so layered, verifiable docs are lacking; score 8.
Setup experience 9/20
PyPI + uvx + stdio gives a short install path and the token is optional thanks to login-based onboarding. But no mainstream-client config examples, platform compatibility notes, or CI/test execution evidence were provided, and the static cap prevents a higher score; score 9.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

  • Developers and AI agents needing programmatic interaction with 3D assets during creation
  • Projects requiring editable, animatable assets with clear part boundaries and high-fidelity geometry
  • Seamless workflow from text or image prompts to structured GLB output

Not for

  • Non-technical users seeking simple opaque meshes
  • Projects requiring offline, local 3D generation without reliance on hosted services
  • Projects needing precise CAD solids (still based on Blender's mesh approach)

Required permissions

  • Network access: communicates with the hosted generation backend to generate 3D assets
  • File system: ability to write generated GLBs and scripts to disk (depending on client)

Risks and side effects

  • API key for hosted service could be leaked, leading to unauthorized use or charges
  • Generated content may inadvertently contain biased or inappropriate content as reflected in prompts or training data
  • Dependence on hosted service availability; outages can affect functionality
  • Generated scripts should be reviewed carefully before use in automated workflows to avoid unintended code execution

Setup

Before you start

Runtime：Python · Hosted (no local runtime)

NOVA3D_TOKEN optionalsecret Advanced/manual fallback API key from https://app.nova3d.xyz/api-key; preferred onboarding is browser sign-in via nova3d_login, so this is optional.
Other optional settings (1)
NOVA3D_APP_URL optional App URL used for conversation links, e.g. https://app.nova3d.xyz or local http://127.0.0.1:5555; optional.

Install the package via pip: pip install nova3d-mcp (version 0.4.0). Optionally set the API key using environment variable NOVA3D_TOKEN (from https://app.nova3d.xyz/api-key) or override the app URL with NOVA3D_APP_URL (for conversation links, default https://app.nova3d.xyz).

Check that it works

After installing nova3d-mcp via uvx, confirm Nova3D tools such as nova3d_login appear in your client's tool list, then generate a simple 3D asset and check the returned preview URL opens.

Troubleshooting

  1. If login fails, check the `NOVA3D_TOKEN` environment variable is set
  2. Verify `NOVA3D_APP_URL` is correct and the app/service is reachable
  3. Confirm the MCP server is accessible over stdio and Python environment is correct
  4. Check the GitHub repository's issues section for known problems

Things to try

Once connected, you can ask your AI assistant things like:

  • Generate a washing machine with detailed internal mechanics as a 3D model
  • Create a GLB model with named, separately addressable parts from this reference image
  • Add rotatable joints at the pivots of the robot arm you just generated
  • Retexture the previously generated drone and export a GLB with its UV atlas

Use cases

Generate structured 3D models with named, editable parts
Integrate AI agents into 3D content generation pipelines via MCP
Create animatable, articulated assets with joints (e.g., robot arms)

Supported clients

Claude Desktop

Listed from the project's documentation, not tested by this site.

Overview

Nova3D is a 3D generation tool by RareSense that turns text or reference images into structured, part-aware 3D assets. It generates Blender-native Python code that compiles to a GLB with named, separately addressable parts, an assembly hierarchy, joint pivots, and PBR materials. Unlike traditional diffusion-based mesh generators, Nova3D treats assets as programs, not baked meshes, preserving full part hierarchy and editability. This MCP server acts as an interface for AI agents, providing outputs like named-part GLB, preview URLs, and Blender scripts.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision 2299e02d57e5 Data synced 2026-10-12 Read the FMRS scoring method