Server has 12 well-named tools with complete input schemas and descriptions. Naming follows verb_noun convention (create_wallet, list_wallets, sign_transaction). All tools have descriptions 50-200 chars, meeting LLM-optimization baseline. Input schemas use proper JSON Schema with types and field descriptions. However, output schemas are not documented, LLMs cannot infer what fields to expect from responses. Error handling is minimal; no recovery guidance or actionable error messages. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications. Missing pagination support for list operations. No idempotency guarantees documented for write operations.

• Document output schemas for all tools. Example: 'Returns {wallet_id: string, name: string, created_at: ISO8601, accounts: [{account_id, address, curve_type}]}'
• Add pagination to list_wallets and list_wallet_accounts: include limit (default 20, max 100), offset/cursor parameters, and return total_count in response.
• Add tool annotations via FastMCP: @mcp.tool(description='...', hints={'readOnly': true}) for list/get tools, hints={'destructive': true} for delete/remove tools.
• Implement error handling with actionable messages. Example: 'Wallet not found (ID: xyz). Call list_wallets() to see available wallets.' Categorize errors as retryable (network timeout) vs fatal (invalid credentials).
• Move api_public_key and api_private_key from initialize_client parameters to environment variables (TURNKEY_API_PUBLIC_KEY, TURNKEY_API_PRIVATE_KEY). Accept only organization_id as a parameter.

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

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