This server provides 6 blockchain query tools with consistent naming (all verb_noun format: get-*) and complete input schemas. However, descriptions are minimal (14-82 chars, below the 50-200 char optimal range), parameter descriptions are generic or missing, and output schemas are not documented. All tools are read-only, which limits complexity, but the implementation lacks LLM-optimization guidance for parameter selection and result interpretation.

• Expand tool descriptions to 80-150 chars each following the WHAT/WHEN/RETURN formula. Example: 'Get the native MON token balance (in wei) for an address on Monad testnet. Use this to check account funds before operations that require a minimum balance. Returns balance as a string integer to avoid precision loss.'
• Add parameter constraints and format guidance to all parameter descriptions. For 'address' parameters: 'Ethereum-format address (0x..., 40 hex chars, case-insensitive).' For 'blockNumber': 'Either the string "latest" or a decimal integer >= 0. If invalid, returns error code INVALID_BLOCK_NUMBER.'
• Document output schemas in tool descriptions or a schema comment block. Example for get-mon-balance: 'Returns {balance_wei: string, formatted_balance: string}. balance_wei is the exact integer value in wei (1 MON = 10^18 wei); formatted_balance is a human-readable string.'
• Add error handling guidance per tool. Example: 'If address is invalid, returns {error: "INVALID_ADDRESS", message: "Address must be 0x... format. Received: {address}"}. Retry with a valid Ethereum address or search_addresses() for available accounts.'
• Declare get-block's blockNumber as an enum in the JSON Schema: blockNumber oneOf [{const: 'latest'}, {type: 'string', pattern: '^[0-9]+$'}]. This prevents LLMs from passing invalid values like 'recent' or 'block_10'.
• Add examples or hints for common patterns. E.g., 'To check if an account has sufficient balance for a transaction, call get-mon-balance first, then compare the returned balance_wei to the transaction cost.' This guides multi-step workflows.

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

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