Jupyter MCP Server is an MCP server developed for AI to connect and manage Jupyter Notebooks in real-time. It exposes a rich set of tools for notebook interaction, organized into server-management commands (list_files, list_kernels), multi-notebook operations (use_notebook, list_notebooks, restart_notebook, unuse_notebook, read_notebook), and cell-level actions (read_cell, insert_cell, delete_cell, overwrite_cell_source, execute_cell, insert_execute_code_cell, execute_code). It also provides JupyterLab integration (notebook_run-all-cells) for a seamless UI experience. The server supports multimodal outputs and can execute cells with timeouts, returning images and other content. It emphasizes context-aware interactions by considering the entire notebook context to make prompts and actions more relevant. It is MCP-compatible, enabling easy integration with any MCP client like Claude Desktop, Cursor, Windsurf, and more. It works with various Jupyter deployments (local, JupyterHub) and with Datalayer-hosted Notebooks. Deployment options include uvx for quick start and Docker for production. Documentation covers client configuration, token handling, and networking tips, along with best practices and prompt usage guidance (e.g., jupyter-cite).