Jupyter Kernel State

Exposes live Jupyter kernel state — variables, dataframe summaries, plots, and tracebacks — to AI coding agents through MCP tools.

Bridges AI coding agents with running Jupyter kernels, surfacing live variable values, dataframe summaries, recent cells, traceback information, and base64-encoded plot captures rather than raw notebook JSON. Provides approximately ten tools for kernel introspection and cell operations, filling the gap left by cell-operation-only Jupyter MCP servers. Supports two modes: server mode (connects to an existing Jupyter instance) and standalone mode (spawns its own kernel). Implemented in Python; pre-alpha and installed from source.