Banking MCP server has 7 tools with basic schemas and descriptions, but significant gaps prevent production use. All tools have input schemas with type definitions and descriptions present, meeting minimum requirements. However, descriptions are uniformly terse (10-20 chars), falling below the 50-200 char baseline for LLM-optimized tools. No output schemas documented. Error handling is absent, tools return raw results with no recovery guidance. Parameter descriptions lack format constraints, ranges, or validation rules. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Missing idempotency guarantees on WRITE operations (deposit, withdraw, open_account). No pagination support on list_accounts or list_transactions despite potential for large result sets.

• Expand all tool descriptions to 50-200 characters. Example: 'deposit' → 'Deposit funds into an account. Requires valid account_number and positive amount. Returns updated balance and transaction ID. Idempotent: same input produces same result.'
• Document output schemas for all tools. Example: open_account returns {account_id: string, account_holder: string, balance: number, created_at: ISO8601}. This enables agents to chain tools and extract required fields.
• Add pagination to list_accounts and list_transactions: accept limit (1-100, default 20) and offset parameters; return {items: [...], total: number, has_more: boolean}.
• Add tool annotations via MCP: mark deposit/withdraw/open_account with destructiveHint=true; mark check_balance/list_accounts with readOnlyHint=true. This signals to LLMs which tools have side effects.
• Implement idempotency for WRITE operations: deposit and withdraw should accept an optional idempotency_key parameter (UUID). Return the same result if called twice with the same key, preventing duplicate charges.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry