AtomisticSkills plugin
Plugin with 130-plus skills for atomistic materials research: relaxation, dynamics, DFT and structure databases through MCP.
- 187 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Calculations run heavy code and download data from external databases on your machine.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/atomisticskillsDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Calculations run heavy code and download data from external databases on your machine.
This is third-party code. Review the repository files before installing.
What it does
AtomisticSkills is a framework for atomistic materials research where an agent composes tasks from three levels: tools, skills and workflows. Tools are strict Python functions behind MCP servers: structure relaxation, NVT, NPT and NVE molecular dynamics, Monte Carlo simulation, DFT input preparation and output parsing, machine-learning potentials. Skills are flexible tutorials that combine calls: MLIP benchmarking, diffusion coefficients, thermodynamic stability, docking with AutoDock Vina and gas sorption isotherms by GCMC. The `skills/` directory holds more than a hundred skills with `SKILL.md`, helper scripts and examples, and `.agents/workflows` contains high-level research roadmaps. The Claude Code plugin installs skills and MCP servers preconfigured, and Codex, Cursor, Gemini, Windsurf and Antigravity use the `configure_mcp.py` script. Calculations run through the `venv/run` launcher in three `uv` environments.
Who it is for. Researchers in computational chemistry and materials science who want an agent to run calculations from ready methods instead of writing a script per task.
Good fit when
- The accuracy of machine-learning potentials is compared on a labeled set with a parity plot.
- A diffusion coefficient and activation energy are computed from a molecular dynamics trajectory.
- A ligand is docked into a protein with AutoDock Vina.
- An agent builds a cluster expansion and runs Monte Carlo from a ready recipe.
Not a fit when
- The workstation runs macOS or Windows, while Linux x86_64 and aarch64 are listed.
- The task lies outside materials science and computational chemistry.
- Calculations are wanted without installing `uv` environments and heavy dependencies.
Example request
Compute the lithium diffusion coefficient from a molecular dynamics trajectory and show the activation energy with a plot.Limitations
Only Linux on x86_64 or aarch64 is listed, glibc 2.28 or newer is needed, and the aarch64 GPU stack needs 2.34. First setup runs `venv/run --setup` and downloads the `cpu`, `mlip` and `fairchem` environments, which takes time and disk space. Calculations and external database downloads run from your machine, and a specialist must review the results. The description was checked against the README as of October 2026.
How to disable. Run `claude plugin uninstall atomistic-skills@atomistic-skills`. For a cloned copy delete the directory and the MCP entries added by `configure_mcp.py`.
FAQ
How do tools, skills and workflows differ?
Tools are strict functions behind MCP, skills are flexible multi-call instructions and workflows are multi-stage research roadmaps.
How do I install for Codex or Cursor?
Clone the repository, run `venv/run --setup`, then `python configure_mcp.py --agent codex --scope global` (or `cursor`, `gemini`, `windsurf`).
What runs in containers?
The plugin uses native `uv` when the system qualifies and falls back to prebuilt container images when requirements are not met.
Is there a quality benchmark?
The project has a separate benchmark with a leaderboard and an evaluation repository. Links are at the top of the README.
Is a GPU required?
Not required: the `cpu` environment works without one, while the heavy `mlip` and `fairchem` stacks target GPUs.
Discussion
No questions or reviews yet.
Related
Plugins24 skills for numerical methods and materials science: the agent checks stability, meshes, convergence and cluster runs.
Skills for computational chemistry: ASE, DFT codes, LAMMPS, DeePMD, RDKit, Open Babel and more, built for OpenClaw
Around 200 skills for Claude Code on bioinformatics, chemistry and biostatistics, from RNA-seq analysis to molecular docking
18 skills and a 34-tool MCP server for 3D Gaussian Splatting: paper reading, method comparison, training debugging and rendering.
Skills for power-system studies: how to work with pandapower, OpenDSS, PSS/E and PowerWorld, plus playbooks for fixing violations.
A Claude Code plugin and MCP server: over 1000 scientific tools for biology, chemistry and medicine, data-check commands and routing skills
README in short
Three abstraction levels, 130-plus skills and MCP tools in one plugin for computational chemistry.