Jupyter has been an extensible computing environment since its creation. It is built around a protocol, plugins, a command system that drives the whole UI, modular packages, and kernels that run code to produce rich representations of data. That combination turns out to be a great fit for coding agents: an agent can reshape the environment to fit how you work, drive it through its commands, and run code in kernels to see the results. In this talk, we will see what makes Jupyter uniquely positioned for agentic workflows.