# Spinel
A materials science toolkit for Claude Code. Install once, get access to materials databases, analysis tools, and domain expertise.
## Quick Start
```bash
/plugin marketplace add charlesxjyang/spinel-plugin
/plugin install spinel
```
That's it. Spinel auto-creates a virtual environment at `~/.spinel/venv`, installs dependencies, and checks your API keys on first session. You just need a Materials Project API key:
1. Create free account at https://next-gen.materialsproject.org
2. Dashboard → API Key
3. `export MP_API_KEY=your_key_here`
## What's Included
### MCP Servers
- **Materials Project** — 150,000+ computed inorganic materials
- **AFLOW** — Crystal prototypes, thermodynamic properties (stub)
- **NOMAD** — DFT calculations, experimental datasets (stub)
### Skills
| Skill | What it does |
|---|---|
| `crystal-structure-analysis` | Load, inspect, compare crystal structures (pymatgen) |
| `phase-diagrams` | Convex hulls, stability, Pourbaix diagrams, CALPHAD (pymatgen, pycalphad) |
| `electronic-structure` | Band structures, DOS, band gap analysis (pymatgen) |
| `xrd-analysis` | Simulate/match XRD patterns, peak indexing (pymatgen) |
| `instrument-data` | Parse vendor files — Bruker, Gatan, Thermo Fisher, Rigaku (hyperspy, tifffile, jcamp) |
| `materials-screening` | High-throughput property search across databases (mp-api) |
| `battery-analysis` | Cycling data, EIS fitting, electrode voltage prediction, PyBaMM modeling (cellpy, galvani, impedance.py, PyBaMM) |
Each skill includes reference documentation that loads into Claude's context on demand — ensuring Claude uses current APIs, not stale training data.
### Auto-Setup
Spinel uses a `SessionStart` hook to automatically:
- Create `~/.spinel/venv` if it doesn't exist (uses `uv` if available, falls back to `venv`)
- Install missing Python packages
- Check for `MP_API_KEY` and warn if missing
No manual environment management needed.
## Core Packages
| Package | Domain |
|---|---|
| `pymatgen` + `mp-api` | Crystal structures, phase diagrams, electronic structure, diffraction |
| `ase` | DFT workflows, structure building, calculator interfaces (VASP, QE, GPAW) |
| `hyperspy` + `kikuchipy` | Electron microscopy (EELS, EDS, EBSD) |
| `pycalphad` + `scheil` | CALPHAD thermodynamics, solidification |
| `phonopy` | Phonon band structures, thermal properties |
| `matgl` | ML interatomic potentials (M3GNet, CHGNet) |
| `diffpy.structure` | Pair distribution function analysis |
| `Dans_Diffraction` | Multi-radiation diffraction simulation |
| `cellpy` + `galvani` | Potentiostat data (Biologic .mpr, Arbin .res, Maccor) |
| `impedance.py` | EIS equivalent circuit fitting |
| `PyBaMM` | Physics-based battery modeling (SPM, DFN) |
## Usage
Just ask Claude naturally:
- "Find stable perovskite oxides with band gaps between 1.5 and 3 eV"
- "Parse this .mpr file and plot the voltage profiles and coulombic efficiency"
- "What's the theoretical capacity and average voltage of LiMnO2 as a cathode?"
- "Fit this EIS data to a Randles circuit and extract the charge transfer resistance"
- "Simulate a 1C discharge of an NMC cell using PyBaMM"
- "Parse this .dm4 file from our TEM and extract the EELS spectrum"
- "Build a CALPHAD phase diagram for the Al-Cu-Mg system"
## Architecture
```
spinel/
├── .claude-plugin/plugin.json
├── .mcp.json # Materials Project + AFLOW + NOMAD
├── pyproject.toml
├── skills/
│ ├── crystal-structure-analysis/
│ │ ├── SKILL.md
│ │ └── references/pymatgen-api.md
│ ├── phase-diagrams/
│ │ ├── SKILL.md
│ │ └── references/
│ ├── electronic-structure/
│ │ ├── SKILL.md
│ │ └── references/
│ ├── xrd-analysis/
│ │ ├── SKILL.md
│ │ └── references/
│ ├── instrument-data/
│ │ ├── SKILL.md
│ │ └── references/
│ ├── materials-screening/
│ │ ├── SKILL.md
│ │ └── references/mp-api-guide.md
│ └── battery-analysis/
│ ├── SKILL.md
│ └── references/
├── scripts/
│ ├── setup.py # Auto-setup (SessionStart hook)
│ └── validate_structure.py # Structure validation
├── servers/
│ ├── materials_project_server.py # Implemented
│ ├── aflow_server.py # Stub
│ └── nomad_server.py # Stub
└── hooks/
└── hooks.json # SessionStart auto-setup
```
## Contributing
We need domain experts to review and improve skill files — especially the "Critical Rules" and "Common Pitfalls" sections, which encode the tribal knowledge that makes the difference between useful and broken AI-assisted analysis.
## License
MIT
Lire le README source