Skip to main content
Open format on GitHub. Compute on demand.

Reproduce any materials experiment. Exactly.

openmaterials is the open commons that defines how an experiment is recorded, node to node. MaterialsCodeGraph is the engine that runs it: real GPU compute, exact reproduction, priced before you commit.

See pricing Launch App

Two platforms. One experiment.

An experiment is a path from one node to another: a starting material, a process, a result. openmaterials defines that format in the open. MaterialsCodeGraph is the engine that executes it.

The open commons

openmaterials

An open source card catalog for materials experiments. Every card is git-native, human-readable, and defines a node-to-node path: what you start with, what you do to it, what you get. Anyone can read it, fork it, or improve it.

  • Open source, community-improved format
  • Git-native experiment cards, free to browse
  • A "Run on MaterialsCodeGraph" button on every computable card
The executable engine

MaterialsCodeGraph

Where a card becomes a run. MaterialsCodeGraph hosts the heavy files, runs the compute on real GPUs, and reproduces the exact experiment an openmaterials card describes, or a new one you define yourself.

  • Real GPU compute, quantum-accurate phonon transport
  • Cost shown in tokens and USD, approved before it runs
  • Every result links back to its openmaterials card

Three loops. One flywheel.

Improving the open map costs nothing. Sharing what you find costs nothing. Running new compute is the only thing you pay for, and you approve the cost first.

1. Improve the map

Free and open source. Add or correct an experiment card on openmaterials, the shared format everyone reads from.

FREE

2. Share a result

Every run and every study has one permanent link with a public/private toggle: flip it public and anyone can view, flip it private and the same link goes dark, free either way, and the link never changes. Study pages stay live, updating as new runs land. Or publish a finished run as an immutable, indexable, citable public page, no login and no token, safe to put in a paper. Either way it cross-links back to its card on openmaterials for anyone who wants to reproduce it.

3. Run with tokens

Launch a new simulation or replay an existing openmaterials lineage on MaterialsCodeGraph. See the token and USD cost, then approve it before it runs.

Open Format and Open Provenance

The format is open.
openmaterials, not us.

The definition of what an experiment is, the node-to-node card format, lives on openmaterials, a public good, open source and free to browse. MaterialsCodeGraph is one engine that reads it; it is not the format's owner.

  • Browse and fork the format on GitHub
  • Contribute or correct an experiment card
  • Free to read, no account required

Provenance you can follow.
Not a black box.

Every run on MaterialsCodeGraph links back to the exact openmaterials card, code commit, and input files behind it. That link is open in every shared run, not gated behind an enterprise tier.

  • Graph-based execution history
  • Every result cross-links to its source card
  • Reproduce it yourself, not just trust it
Born at Da Vinci Labs

Built by the team that wrote the code.

MaterialsCodeGraph began as the internal physics engine for Da Vinci Labs, a deep-tech consultancy. Our founder was the lead developer of kALDo, the industry-standard open source framework for phonon transport.

We built this platform to solve our own consulting bottlenecks. Now, it's open to the world.

Open source format, read by anyone, for free
Cost shown in tokens and USD before a run starts
Every result cross-links back to its openmaterials card
The full experiment, rendered: migration barriers, switching-voltage statistics, spectra, one shareable link
Drag and drop your own input files when creating a simulation, up to 50 GB
Upload and share your data with a link before running anything: no simulation required

Supported Integrations

Quantum Espresso Electronic Structure
LAMMPS Molecular Dynamics
kALDo Thermal Transport
MACE, ORB, MatterSim ML Potentials
Compatible with OpenRoad / Chipyard

Free to learn and share. Pay only for commercial compute.

Students and academic researchers get storage and compute free up to a generous limit. Commercial use is billed as a single token figure covering both, and you approve the quote before anything runs.

Sharing an experiment on openmaterials is always free.

Students & Academia
Free

The lane to learn and publish. Storage and compute are free up to a generous limit, so you can run real experiments without a budget.

  • Storage free up to a generous limit
  • Compute free up to a generous limit
  • Sharing on openmaterials always free
PAY PER TOKEN
Commercial
Pay per token

Run real GPU compute and store your results, billed as one combined token figure covering both. You approve the quote before anything runs.

  • One token figure for storage and compute
  • Quote shown before you commit, every run
  • Sharing on openmaterials still always free
Enterprise
Talk to us

Dedicated capacity, private workspaces, or a custom integration. Tell us what you're trying to reproduce.

Reproduce your first experiment.

Browse lineages on openmaterials, or run simulations on MaterialsCodeGraph now. Join the waitlist below for updates on what's next.

For enterprise deployments or questions: hello@materialscodegraph.com