A browser-based platform for the full 3D molecular generative design pipeline: run pretrained MolCraftDiffusion models, curate and enrich datasets, and explore chemical space — all from one web interface, no scripting required.
| De-novo & property-guided generation | DDPM/DDIM sampling with classifier-free guidance toward per-property targets, run directly from the browser |
| Structure-guided generation | Inpaint or outpaint from a reference .xyz scaffold, with tunable denoising/constraint strength |
| Model training | Configure, queue, monitor, and export MolCraftDiff training jobs from a form or an imported YAML |
| Multi-source data curation | Stage and compile CSV+XYZ, ASE .db, and generation-job outputs into one dataset |
| Analysis pipeline | Async jobs for validity/connectivity checks, XTB properties and geometry optimization, featurization, dimensionality reduction, and property prediction |
| Linked visualization | 2D/3D scatter, histograms, and a 3D molecule viewer sharing one selection state, GPU-rendered via deck.gl |
| Plug-in tools | Wire in external property predictors by dropping a manifest.json + runner.py — no backend changes needed |
git clone https://github.com/pregHosh/AutomaticMolCraftt && cd AutomaticMolCraftt
conda env create -f environment.yml # macOS (Apple Silicon): environment-macos.yml
conda activate molcraft
automolcraft doctor # checks xTB, MolCraftDiffusion, GPU, models; says what's missingThe environment file installs xTB, OpenBabel, Node.js, MolCraftDiffusion at the pinned commit
(b79e8aad…, version 1.12.0), the analysis-tool packages, and this app (which provides the
automolcraft command). Download pretrained models from Hugging Face
into models/, then launch:
automolcraft serve # builds the frontend on first run, opens http://127.0.0.1:8000Options: --port 9000, --host 0.0.0.0, --dev (Vite hot reload), --rebuild, --no-browser.
Manual install — the step-by-step route still works and launches with ./dev.sh:
conda create -n molcraft python=3.11 -y && conda activate molcraft
conda install -c conda-forge xtb==6.7.1 openbabel -y
MOLCRAFT_REF=b79e8aadc85f7047fbd9a70d1c41ea3aba0fc0a7
pip install "molcraftdiffusion[gpu] @ git+https://github.com/pregHosh/MolCraftDiffusion@${MOLCRAFT_REF}" \
--find-links https://data.pyg.org/whl/torch-2.6.0+cu124.html # or [cpu] with the CPU torch index
pip install -r webapp/database-explorer-lite/backend/requirements.txt
cp webapp/database-explorer-lite/.env.example webapp/database-explorer-lite/.env
./dev.shSee the installation guide for macOS notes, CPU-only Linux, environment variables, and all launch options.
The WebUI has seven tabs:
| Tab | Purpose |
|---|---|
| Visualization | Explore the compiled dataset with linked plots, filters, and the 3D viewer |
| Management | Register, compile, filter, and export datasets |
| 3D molecule generation | De-novo or property-guided generation |
| Structure-directed generation | Inpaint/outpaint from a reference structure |
| Analysis tools | Queue analysis jobs or build multi-step workflows |
| Model training | Configure, queue, and monitor MolCraftDiff training jobs |
| Plug-in tools | Run locally installed external tools |
See the tutorials for a full walkthrough of each tab.
- Installation
- Quick Start
- UI Overview
- Analysis Tools
- Model Training
- FAQ
- MolCraftDiffusion Repository & Docs · Docs
If you use AutomaticMolCraft in your research, please cite:
@article{worakul_automaticmolcraft_2026,
title = {{AutomaticMolCraft}: {A} {Web} {Interface} for {3D} {Molecular} {Generation}, {Dataset} {Curation}, and {Structure}–{Property} {Analysis}},
url = {https://chemrxiv.org/doi/abs/10.26434/chemrxiv.15008080/v1},
doi = {10.26434/chemrxiv.15008080/v1},
publisher = {American Chemical Society (ACS)},
author = {Worakul, Thanapat and Hernandez Cuellar, Osvaldo and Corminboeuf, Clémence},
month = aug,
year = {2026},
}If you use MolCraftDiffusion, the generative engine this app is built on, please cite:
Modular Framework for 3D Molecular Generation in Computational Chemistry Applications
@article{worakul_modular_2026,
title = {Modular {Framework} for {3D} {Molecular} {Generation} in {Computational} {Chemistry} {Applications}},
copyright = {https://creativecommons.org/licenses/by/4.0/},
issn = {0002-7863, 1520-5126},
url = {https://pubs.acs.org/doi/10.1021/jacs.5c19960},
doi = {10.1021/jacs.5c19960},
language = {en},
urldate = {2026-06-24},
journal = {Journal of the American Chemical Society},
author = {Worakul, Thanapat and Azzouzi, Mohammed and Wodrich, Matthew D. and Corminboeuf, Clémence},
month = jun,
year = {2026},
pages = {jacs.5c19960},
}
A Diffusion Framework for Geometrically Valid and Practically Viable 3D Molecular Generation
@article{worakul_diffusion_2026,
title = {A {Diffusion} {Framework} for {Geometrically} {Valid} and {Practically} {Viable} {3D} {Molecular} {Generation}},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15005231/v1},
doi = {10.26434/chemrxiv.15005231/v1},
publisher = {American Chemical Society (ACS)},
author = {Worakul, Thanapat and Corminboeuf, Clémence},
month = jun,
year = {2026},
}This project is released under the MIT License.