The AI IDE forJupyter notebooks.
The notebook has barely changed in decades. Clusy turns it into an AI-native development environment. Work with an agent, branch experiments, and run anything on scalable compute.
Backed byFounders, Inc.
Three things a notebook can’t do on its own.
Start from an empty notebook and one sentence. It finds the data and writes the cells, and when the first one throws it declines the fix everyone reaches for, because that fix would have quietly biased the sample against exactly the stars the question was about.
Loved by researchers and engineers
“I haven’t seen many tools that can handle the full LLM fine-tuning workflow end-to-end — dataset discovery, preprocessing, sandboxed training, parallel experiments — all from a single prompt.”
“With a single natural language prompt, it completed a full ML workflow — finding public datasets, reproducing a model from a paper, and running experiments. Clusy has the potential to redefine the workflows of many researchers.”
“I was very impressed by the interface and the end-to-end experience that I was lacking on other platforms.”
“We’ve been using Clusy to post-train open-weight models for our coding agent system, and it’s been working really well for us. It has made the workflow much faster and saved our team a lot of time compared to doing everything manually.”
“It autonomously fetched the data from a public OSF dataset, handled it perfectly, and pointed out edge cases without me instructing it to. I really loved the end-to-end reasoning — I only needed to minimally guide it.”
“Fine-tuned a model for widget extraction without touching a single env config. The notebook it handed back was clean enough that I pulled it straight into the repo.”
“Clusy has been really useful and I believe the platform has a lot of potential. I love the fact that the agent is there for you in the entire process.”
“Clusy can handle the entire pipeline: fetching relevant training data, cleaning it, trying different model architectures, and training the model on the cloud. It’s amazing.”
“I used Clusy for idea validation and market research, and it produced very insightful results, finding exactly what I was looking for. I can see a variety of use cases for engineers and researchers.”
“I really enjoyed the platform layout and look :)) it looks very easy to manage a project.”
Frequently asked questions
What is Clusy?
Clusy is an AI IDE for Jupyter notebooks. You describe the outcome you want in plain language, and the Clusy agent plans the work, writes and runs notebook cells on cloud CPUs or GPUs, and returns results you can inspect, edit, and re-run. Every step stays visible in a real notebook, so you keep full code-level control.
How is Clusy different from Jupyter or Google Colab?
Jupyter and Colab give you an empty editor and leave the work to you. Clusy is the IDE around that notebook: an agent works inside it with you, sourcing and inspecting data, choosing architecture and compute, and executing end to end. It also adds two things a single notebook cannot do on its own: branching, so you can fork and race experiments in parallel and compare them side by side, and snapshot sharing, so any run becomes a stable link others can open and fork. The output is still a real notebook you can read, edit, and export.
Can I bring my existing .ipynb files, and take them out again?
Yes, in both directions. Import a .ipynb straight into a Clusy notebook, and export any notebook back to .ipynb whenever you want. From a shared notebook you can also publish to Hugging Face, Kaggle or GitHub in one click, or download it and open it in Google Colab. Nothing you build is locked in.
Which AI models can I use in Clusy?
The Free plan includes the Auto model, which routes every request to the best value model for fast, low-cost everyday work. Plus adds DeepSeek V4 Flash and V4 Pro, Pro adds every open model such as Kimi, Qwen and GLM, and Max unlocks every model, including Anthropic Claude (Sonnet and Opus) and OpenAI GPT.
What GPUs and compute does Clusy provide?
Every plan runs on managed cloud sandboxes. Free starts with an 8 vCPU / 8 GB RAM CPU sandbox. Plus adds an entry T4 GPU with 16 GB VRAM and 16 GB sandbox RAM, Pro adds the L4 and A10 up to 24 GB VRAM with 32 GB sandbox RAM, and Max scales up to A100, H100 and H200 GPUs up to 141 GB VRAM with up to 128 GB sandbox RAM.
Can Clusy connect to my own data?
Yes. You can upload files, pull public datasets from sources like Hugging Face, or connect directly to your Databricks or Snowflake warehouse and work against live tables from the notebook.
How much storage does each Clusy plan include?
Storage on Clusy is checkpoint storage: the saved kernel state a project keeps so you can come back and pick up where you left off. Notebooks, files and uploads are not what fills it. Free includes 5 GB, Plus 30 GB, Pro 100 GB, and Max 512 GB at the 10x usage level or 2 TB at 30x. On the paid plans an idle sandbox stays paused; on Free it is released after 3 days, and the project and its saved state stay restorable.
Is there a free plan?
Yes. Clusy is free to start: the Free plan includes the Auto model on a CPU sandbox with no credit card required. Paid plans start at $12/month (Plus), with Pro at $30/month and Max at $90 or $200/month for 10x or 30x the Plus usage, all with usage-based allowances and optional pay-as-you-go.