MatCreator

An agentic AI platform for computational materials science.
Build workflows, generate structures, run simulations, and continuously acquire new skills.

Documentation · Quick Start · Launch

---

✨ Features

• 🧠 Agentic workflow planning using graph-based execution
• 🔬 Materials science focused, from crystal generation to MLFF workflows
• 🧩 Skill-based architecture — capabilities are modular and extensible
• 📚 Self-improving knowledge graph that accumulates experience over time
• 🌐 Modern Web UI with graph visualization and artifact management
• ⚡ CLI + API + Web interfaces

---

Quick Start

Installation

git clone https://github.com/AI4MS/MatCreator.git
cd MatCreator

pip install uv
uv venv .venv --python 3.12
source .venv/bin/activate

uv pip install -e .

Configure your LLM credentials through the MatCreator CLI

matcreator config set llm.model=openai/qwen3-plus
matcreator config set llm.api_key=your-api-key
matcreator config set llm.base_url=https://api.example.com/v1

Start an interactive session from the workspace you want MatCreator to use

matcreator chat --workspace .

[Optional] Install official skills (SkillForge)

MatCreator can install corporation-maintained official skills from the SkillForge repository managed by the MatCreator team. These skills are installed into ~/.matcreator/skills/official and loaded alongside built-in and custom skills.

Notice: Only authenticated users with repository access can install official skills from SkillForge.

matcreator install official --repo [email protected]:AI4MS/SkillForge.git --ref devel

If your repository requires authentication, configure your Git SSH keys or credential helper first.

[Optional] Install the Vite frontend

The Vite frontend uses Ketcher and requires Node.js 24.14.1 or later. Install the frontend dependencies with:

cd web/vite-frontend
npm install

Check the installed versions with:

node --version
npx vite --version

---

Launch

Web Interface (Recommended)

bash script/start_matcreator.sh

This starts

• ADK API
• FastAPI backend
• Vite frontend

---

CLI

matcreator chat

matcreator chat --workspace ~/materials-project

By default, matcreator chat starts in Flash mode for fast direct interaction. Use matcreator chat --plan when you want the full planning and graph-execution workflow.

---

Documentation

Full installation, configuration, and deployment documentation is available at ai4ms.github.io/MatCreator.

---

Vision

MatCreator aims to become a continually evolving AI scientist for computational materials research by combining

• Large language models
• Scientific software
• Modular skills
• Long-term knowledge accumulation

into a unified agentic system.

---

License

Apache 2.0