Critical gaps across naming, descriptions, and schemas. Only 1 of 17 tools has a complete, visible schema (get_scraped_data). Most tool definitions are inferred from service files rather than explicitly registered in the MCP server code. Descriptions vary from adequate (get_scraped_data: 156 chars) to minimal (verify_zkproof, verify_teeattest, subscribe_stream: empty). Parameter descriptions are sparse or missing entirely. No output schemas documented. Error handling is minimal. The server exposes payment-gated APIs but lacks guidance on cost, retries, or failure recovery.

• Explicitly register all 17 tools in src/mcp-server.ts with full schemas and descriptions. Do not rely on inferred definitions from service files. Use the same pattern as get_scraped_data for all tools.
• Add non-empty descriptions to verify_zkproof, verify_teeattest, and subscribe_stream. Each must explain WHAT the tool does, WHEN to use it, and any prerequisites.
• Document output schemas for all tools. Include field names, types, and descriptions. E.g., get_quote should return {routeId: string, inputAmount: string, outputAmount: string, priceImpact: number, ...}.
• Add error recovery guidance to payment-gated tools. Document: (1) which errors are retryable, (2) cost implications of retries, (3) fallback strategies if payment fails. E.g., 'If payment fails, the call is not charged. Retry after checking wallet balance.'
• Expand parameter descriptions for infer_* and execute_route tools. For each parameter, state the expected format, valid range, and any constraints. E.g., 'taxonomy: Classification taxonomy (string, max 100 chars, e.g. "sentiment", "intent", "entity_type")' instead of bare 'Classification taxonomy'.