Server has 5 tools with basic schemas and descriptions, but significant gaps prevent production use. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization) and input schemas are present but minimal. Parameter descriptions are entirely absent, LLMs cannot infer what 'contractId', 'side', 'type', 'amount', or 'limitPrice' mean without explicit guidance. The placeOrder tool (IRREVERSIBLE risk) lacks confirmation/dry-run support and has no error recovery guidance. Output schemas are undocumented. No tool annotations (readOnlyHint/destructiveHint) despite clear risk differentiation. Naming is verb-forward (getMarketData, placeOrder) but descriptions are too terse to guide LLM selection.

• Add parameter descriptions to all tools. For 'contractId': 'The unique identifier of the contract (e.g., BTC-USD). Use getContractDetails or list available contracts first.' For 'side': 'Order direction: buy or sell.' For 'type': 'Order type: market (execute immediately at current price) or limit (execute only at specified price).' For 'amount': 'Quantity to trade (numeric, e.g., 0.5 for 0.5 BTC). Must be positive.' For 'limitPrice': 'Price limit for limit orders (numeric, e.g., 45000.50). Required if type=limit.'
• Expand tool descriptions to 100-150 chars. Example: 'Get real-time market data (bid/ask/last price, volume) for a contract. Use this to check current prices before placing orders. Returns ticker data updated every second.'
• Add tool annotations: readOnlyHint=true for getMarketData/getContractDetails/getAccountInformation/estimateOrder; destructiveHint=true for placeOrder. This signals to LLMs which tools are safe to retry.
• Implement dry-run support for placeOrder: add optional 'dryRun' boolean parameter (default false). When true, return estimated execution price and fees without executing. Require explicit confirmation for live trades.
• Document output schemas in tool descriptions or as a separate schema field. Example for getMarketData: 'Returns {bid: number, ask: number, last: number, volume: number, timestamp: ISO8601}'. This enables LLMs to extract and chain data correctly.