AutoGL is an open-source AutoML framework and toolkit dedicated to machine learning on graph data, developed by THUML Lab at Tsinghua University. The GitHub project has around 1,100 stars and is built on PyTorch and PyTorch Geometric. It was still receiving updates in late 2025.
Graph machine learning involves many moving parts — feature engineering, model selection, architecture search and hyperparameter tuning — and AutoGL automates that pipeline end to end for graph tasks. It provides automated feature engineering, model training and selection, neural architecture search and hyperparameter optimisation for problems such as node classification, link prediction and graph classification. The framework is modular: researchers can plug in their own models, search spaces and solvers, while practitioners can run the automated pipeline on benchmark or custom graph datasets with a few lines of code. AutoGL comes from the team behind several widely used graph learning libraries, and it includes documentation, examples and support for standard datasets to make automated graph learning reproducible.
AutoGL supports both dataset-level automation — finding a strong model for a whole graph dataset — and finer-grained search over architectures and hyperparameters. Its design mirrors the lab’s broader AutoML research, and the documentation includes tutorials for the main benchmark tasks.
AutoGL is fully free and open source under the Apache-2.0 license. There is no commercial edition or hosted service; you run experiments on your own infrastructure.
Graph ML researchers and data scientists who want to automate model and hyperparameter search for graph tasks, and engineers applying GNNs to networks, molecules, recommendations or fraud detection.