FetchMux MCP defines 2 tools with clear verb-noun naming (search_web, preview_search_route) and reasonable descriptions (60-80 chars). Both tools have identical input schemas with 3 parameters (query, budget, deadline). However, critical gaps exist: (1) parameter descriptions are minimal or missing detail about budget structure and deadline units; (2) output schemas are not documented, LLMs cannot predict response structure for chaining; (3) no error handling guidance; (4) budget and deadline parameters lack type definitions and constraints (e.g., is deadline milliseconds or seconds? what fields does budget object contain?). Tool descriptions are adequate but lack WHEN/WHY context.

• Document the complete output schema for both tools. Example: search_web returns {results: [{title, url, snippet, provider, cost, latency}], receipt: {route_reason, provider_selected, budget_remaining}}. This enables LLMs to extract and chain results.
• Expand budget parameter description to specify structure: 'budget: {max_cost_cents?: number, max_latency_ms?: number, preferred_freshness_hours?: number}'. Include which fields are optional and their valid ranges.
• Add deadline parameter description: 'deadline: number (milliseconds from now, 1-3600000). Request will be cancelled if not completed by this time.'
• Add error handling section to tool descriptions: 'Returns error if: (1) budget exceeded, try with higher max_cost_cents; (2) deadline passed, increase deadline; (3) no providers available, check provider status via preview_search_route.'
• Clarify tool selection guidance: 'Use search_web to retrieve results. Use preview_search_route first to see which providers are eligible and estimated costs before committing to a search.'