Server has 8 well-named, read-only tools with consistent verb_noun naming (get_*, list_*). All tools have descriptions (avg 110 chars, within baseline 34-392). Input schemas use Zod with type definitions and descriptions for all parameters. However, output schemas are completely undocumented, responses are JSON-stringified with no field documentation, forcing LLMs to infer structure. Error handling is minimal (generic fail() wrapper). No parameter enums despite several accepting constrained values (e.g., currency codes). Missing output schema documentation is a critical gap for composition and chaining.

• Document output schemas for all tools. For list_orders, specify: returns {orders: [{id, amount, currency, status, payment_system, email, created_at, updated_at}], total_count, page, page_size}. For get_balance, specify: returns {balances: [{currency, amount}]}.
• Add enum constraints to currency parameters. Replace free-form 'from'/'to' strings with enum: ['RUB', 'USD', 'EUR', ...] based on Tegro.Money's supported currencies.
• Enhance error handling with recovery guidance. Instead of fail(e), return: {isError: true, content: [{type: 'text', text: 'Shop not found. Verify shop_id or call list_shops() to see available shops.'}]}
• Document pagination limits in tool descriptions. Add: 'Returns up to 50 orders per page. Use page parameter for pagination. Call with page=1 to start.'
• Validate shop_id resolution early. If resolveShopId(shop_id) returns null/undefined, return actionable error: 'shop_id required. Set TEGRO_SHOP_ID env var or pass shop_id parameter.' instead of silent failure.