Jupyter notebooks with serverless compute
Open a project, start a CPU or GPU runtime in one click, and publish the result — no infrastructure to manage.
Runtimes on demand
Boot a CPU or GPU sandbox in one click — from a small CPU to an H100. You pay only while it runs.
Publish & share
Publish any notebook to a public read-only link, or share it live with viewer and editor roles.
Built for agents
An MCP server lets coding agents browse projects, edit cells, and run notebooks on your runtimes.
Local-first files
Projects are plain folders of .ipynb notebooks and files. Upload, download, and export anytime.
Pricing
Free during the beta
Tensor Pad is free while in beta. Usage-based pricing with included monthly credits launches later.