Awesome Gaussian Skills plugin
18 skills and a 34-tool MCP server for 3D Gaussian Splatting: paper reading, method comparison, training debugging and rendering.
- 184 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The MCP server renders and edits scenes on the local machine through Node.js.
- The `setup.sh` install script runs through `curl | bash` from the repository.
Install
- Works in
Manual install
/plugin marketplace add jaccen/Awesome-Gaussian-SkillsThen install `gaussian-research`, `gaussian-engineering` or `gaussian-domain`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The MCP server renders and edits scenes on the local machine through Node.js.
- The `setup.sh` install script runs through `curl | bash` from the repository.
This is third-party code. Review the repository files before installing.
What it does
The repository turns a catalog of 1066 3D Gaussian Splatting methods into an agent toolkit. Eighteen skills live in `skills/` and are grouped into three marketplace plugins: `gaussian-research`, `gaussian-engineering` and `gaussian-domain`. They cover paper reading, method comparison across ten-plus dimensions, venue-tailored experiment planning, detection of 104 known code bug patterns, training debugging, compression and deployment, NeRF to 3DGS migration and CAD conversion. A separate `cg-paper-writing` skill leads a paper from idea to rebuttal. Other skills prepare software patent filings and assemble a team of research agents. The Node.js MCP server works over stdio and exposes 34 tools: 21 core ones such as `import_scene`, `set_camera`, `modify_gaussians` and `render_frame`, plus 13 experimental stubs. An interactive method explorer runs on GitHub Pages.
Who it is for. Computer graphics researchers and engineers who read 3DGS papers, compare methods and prepare publications with an agent.
Good fit when
- You need a fast comparison of 3DGS and 2DGS by rendering formulation and speed.
- Scene training diverges and the agent needs a check against known bug patterns.
- A project is being moved from NeRF to 3DGS following migration steps.
- A paper is being prepared for a conference deadline: skeleton, experiments, figures and a rebuttal.
Not a fit when
- The work is outside Gaussian Splatting: the skills are specialized to this topic.
- You need a ready rendering engine: part of the MCP server consists of stubs.
- You only want a link catalog of papers without tooling.
Example request
Compare 3DGS and 2DGS by rendering formulation and training speed and give conclusions for a scene with reflective surfaces.Limitations
Of the 34 MCP server tools, 13 are experimental stubs and only the 21 core ones are complete. The server is built from `mcp-server/` and launched with `node` and the path to `dist/index.js`. Skills are copied into the agent directory by hand in one batch, while `setup.sh` installs through `curl | bash`, so read it before running. The method database was collected automatically and citation or paper data can lag. Patent filings and paper text come out as drafts that a human must verify.
How to disable. Remove the plugins with `/plugin uninstall`, delete the copied skills from `.claude/` or `.cursor/rules/`, and remove the MCP server entry from the client config.
FAQ
Which plugins are included?
Three: `gaussian-research`, `gaussian-engineering` and `gaussian-domain`. Together they hold 18 skills.
How do I install without plugins?
Clone the repository and copy `skills/*` into `.claude/` for Claude Code or `.cursor/rules/` for Cursor.
What can the MCP server do?
It imports scenes, moves the camera, modifies Gaussians, renders frames and runs a sculpt pipeline. There are 34 tools, 21 of them core.
Does it help write a paper?
Yes, one skill covers the full paper lifecycle: from idea and skeleton through figures and polish to a rebuttal.
Where can I browse the method catalog?
In the interactive explorer on GitHub Pages: search, category filters and sorting by citations.
Discussion
No questions or reviews yet.
Related
PluginsPlugin with 130-plus skills for atomistic materials research: relaxation, dynamics, DFT and structure databases through MCP.
24 skills for numerical methods and materials science: the agent checks stability, meshes, convergence and cluster runs.
The official Graphlit MCP server: an agent collects material from Slack, GitHub, mail and websites into a knowledge base and searches it
A skill for math modeling contests CUMCM, MCM/ICM and the Electrical Cup: 10 stages, three roles, five gates and a LaTeX paper
Skills for power-system studies: how to work with pandapower, OpenDSS, PSS/E and PowerWorld, plus playbooks for fixing violations.
An MCP server for PostgreSQL with several databases at once, table descriptions from the catalog, PostGIS and pgvector hints and query plans
README in short
The project pairs a catalog of 1066 methods in 26 categories with a skill set so that an agent reads papers for the human. Installation for Claude Code, Cursor, OpenClaw and DSH is described, along with the architecture and a roadmap. An online book on spatial and embodied intelligence is included.