Server has 7 tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Most tool descriptions are under 100 chars and lack actionable context. Parameters lack type constraints (enums, ranges). No error handling guidance visible. The 'connect' tool exposes authentication tokens as parameters (critical security issue). Output schemas are not documented. Tool composition is reasonable (single responsibility), but descriptions do not explain WHEN to use each tool or what to expect in responses.

• Move document_token and runtime_token from 'connect' parameters to environment variables (JUPYTER_DOCUMENT_TOKEN, JUPYTER_RUNTIME_TOKEN). Update 'connect' to accept only provider, document_url, document_id, runtime_url, runtime_id. Inject tokens server-side.
• Document output schemas for all tools. E.g., list_cells returns {cells: [{index: int, type: string, source: string, outputs: array}], total: int}. Include this in tool descriptions or a separate schema reference.
• Add parameter constraints: cell_index (integer, 0-based, required), code (string, max 10000 chars, required), limit (integer, 1-100, default 20 for paginated results). Use enums where applicable (e.g., cell_type: 'code'|'markdown'|'raw').
• Enhance error handling: execute_code and execute_cell should return {success: bool, output: string, error: {type: 'syntax'|'timeout'|'kernel_crash', message: string, recovery: string}}. E.g., 'Syntax error on line 5. Check code and retry.' or 'Kernel crashed. Call get_kernel_info to check status.'
• Expand tool descriptions to 100-200 chars, explaining WHEN to use each tool and what to expect. E.g., 'Execute a single cell by index and return its output. Use after list_cells to identify the cell index. Returns stdout, stderr, and display outputs.'