Server has 6 well-structured tools with complete input schemas and descriptions. All tools follow verb_noun naming (calculate, solve_symbolic, solve_numeric, check_units, plot, plot_image). Descriptions are substantive (100-200 chars), explaining WHAT each tool does and WHEN to use it. Parameters are typed with enums where appropriate (operation, method, mode, domain, check). However, output schemas are not documented, LLMs cannot predict return structure. Error handling is absent from visible code; no recovery guidance or actionable error messages. The plot_image tool is marked experimental/test, suggesting incomplete validation. Missing per-parameter descriptions for some optional fields (e.g., 'extra' in solve_symbolic lacks detail on valid keys). No security annotations or permission gates visible.

• Document output schemas for all tools. Example: 'Returns {result: number, units: string, latex: string}' for calculate. Include field types, nesting, and whether arrays are paginated.
• Add error handling with actionable recovery guidance. Example: 'If expression is invalid, return {error: "SyntaxError", message: "Unexpected token at position X", suggestion: "Did you mean...?"}'. Categorize errors as retryable vs user-fixable.
• Expand 'extra' parameter description in solve_symbolic to enumerate valid keys: 'extra: {order: int for diff, bounds: [a,b] for integrate, n: int for series, ...}'.
• Remove or complete plot_image. If experimental, document the expected graduation timeline and known limitations. If production-ready, remove the [EXPERIMENT/TEST FEATURE] label.
• Add tool annotations via FastMCP. Mark all tools with readOnlyHint=true since they do not modify state. This enables smarter client retry logic and clearer semantics.