This is an SDK/framework, not a production MCP server. The three tools shown are example implementations with minimal quality. web_fetch appears twice (likely a duplicate in the listing). Descriptions are extremely terse (16-34 chars, well below the 50-200 char baseline for LLM-optimized tools). Input schemas are present but lack parameter descriptions, only 'url' and 'expression' are named with minimal context. No output schemas documented. No error handling guidance. The tools are READ_ONLY, which is good, but the definitions lack the depth needed for reliable agent invocation.

• Expand tool descriptions to 50-200 chars. Example: 'web_fetch: Fetch and return the full text content of a URL. Use this to retrieve web pages, documentation, or API responses. Returns plain text; fails on 404, timeout (>10s), or non-text content types.'
• Add parameter descriptions with constraints. Example: 'url (string, required): The full HTTP(S) URL to fetch. Must start with http:// or https://. Max 2048 chars. Timeouts after 10 seconds.'
• Document output schemas. Example: 'web_fetch returns {content: string, status: number, headers: object}. calculate returns {result: number, expression: string}.'
• Add error recovery hints. Example: 'If web_fetch fails with 404, the URL may be incorrect. If it times out, the server may be slow, retry after 5 seconds. If it returns non-text, try a different URL.'
• Remove the duplicate web_fetch entry or clarify if it serves a different purpose (e.g., one for client-side, one for server-side).