Aim is an easy-to-use, supercharged open-source experiment tracker for machine learning: it logs your training runs and AI metadata, then gives you a polished web UI to compare runs and a Python API to query everything programmatically. Designed to handle large numbers of runs without slowing down, it positions itself as a faster, lighter alternative to tools like TensorBoard and MLflow. Aim has about 6,300 stars on GitHub and remains actively maintained, with updates through 2026.
Serious model development produces thousands of runs, each with metrics, hyperparameters, images, audio and text samples. Aim stores that metadata in a structured repository you control, and its UI is built for the questions practitioners actually ask: which run had the best loss under these settings, how does this metric evolve across experiments, and what changed between two checkpoints. Integrations cover PyTorch, TensorFlow, Keras, XGBoost, LightGBM and more, and prompt-engineering workflows are supported too, so LLM experiments can be tracked alongside classic training. Because everything is self-hosted and queryable through an API, teams keep full ownership of their experiment data while still getting fast, visual comparisons across very large run collections.
Aim is completely free and open source under the Apache-2.0 license. There is no paid tier; you install it with pip and run it against your own storage.
Aim is for ML engineers, data scientists and research teams who run many experiments and need a fast, visual way to compare them, especially teams that prefer a self-hosted tracker they fully control over a hosted MLOps platform.