IOL MCP server has 8 tools with basic naming and descriptions, but significant gaps in schema completeness and parameter documentation. Tool names follow verb_noun convention (get_*, set_*, cancel_*), which is good. However, most tools lack comprehensive input validation, parameter descriptions are minimal, and output schemas are not documented. The server exposes credentials via parameters (get_token accepts username/password), violating secret-injection patterns. Error handling is present but generic. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels. Average tool score: 52/100.

• Remove username and password from get_token parameters. Instead, require IOL_USERNAME and IOL_PASSWORD environment variables set on the server. If interactive login is needed, implement a separate tool that returns a login URL or uses OAuth flow, not raw credentials.
• Add tool annotations: mark cancel_operation with destructiveHint=true, mark get_* tools with readOnlyHint=true, mark set_active_user with idempotentHint=true. This enables agents to reason about side effects.
• Document output schemas for all tools. For get_operations, specify: returns array of {operation_number, status, country, date, ...}. For get_portfolio, specify: returns {cash, positions: [{symbol, quantity, price}]}. Include field types and descriptions.
• Add pagination to get_operations: accept limit (1-100, default 20) and offset parameters. Return {operations: [...], total_count: N, has_more: bool}. Document the limit in the description.
• Enhance error messages with recovery guidance. E.g., 'No valid token for user. Call set_active_user(username) first, then get_token(username, password) to authenticate.' or 'Token expired. Call get_token() to refresh.'