Clusy

About Clusy

Clusy is an agent-native notebook for machine learning and data science — the pitch is that instead of writing cells, you describe the experiment and it runs every branch. Give it a research question and it finds the data, writes the cells, provisions the compute, runs the work and hands back the branch that succeeded, with the notebook filling in as it goes so you can watch and intervene. You can queue a follow-up while a run is in flight rather than waiting for it to finish. The workflows it advertises end to end are the ones that normally span four tools: dataset discovery, preprocessing, sandboxed training, and parallel experiments, all from a single prompt. Testimonials on the site come from named researchers — a MATS research fellow and OpenAI red-teaming contractor, a Tsinghua postgraduate researcher, an incoming Google engineer, and the co-founder of a coding-agent company using it to post-train open-weight models — describing it fetching data autonomously from public repositories and reproducing a model from a paper. Compute is part of the product rather than something you attach: the free tier gives a CPU sandbox and a fast default model, and higher tiers unlock progressively larger GPUs from T4 and L4 up through A100, H100 and H200, plus access to open-weight models and frontier models including Claude and GPT. The company is backed by Founders, Inc.

Key Features

Clusy Pros & Cons

Tags