Early access

Labeled gameplay data
built for AI training.

Every tick, every input, every event — parsed, labeled, and ready to feed a model. Across games, delivered through the interface your pipeline already speaks.

Datasets

Browse parsed matches across games. Inspect schema, coverage, and sample rows before you commit.

Open the test suite →

Access

REST endpoints for queries. WebSockets for streams. Bulk parquet for training runs.

View access modes →

Notebooks

Coming soon

Hosted Jupyter environment. Point at a dataset and train — no export needed.

Preview notebooks →

Models

Coming soon

A registry for the models you train here. Evaluate, version, and share.

See registry →

Dev environment. Data under ./output is served from a local FastAPI backend on :8000.