The server defines 5 tools with generally clear purposes, but exhibits significant gaps in parameter descriptions, output schema documentation, and error handling guidance. Naming is mostly verb-forward and clear. Tool descriptions are present and reasonably detailed (range 73 - 289 chars), but parameter-level documentation is sparse or missing, and output structures are not documented. The server implements a pragmatic x402 payment fallback but does not guide LLMs on recovery steps when payment fails. Risk declarations are present for all tools (READ_ONLY) but security implications of the payment flow are underexplored in descriptions.

• Document the output schema for each paid tool (cerebrus_pulse, cerebrus_sentiment, cerebrus_funding). Include field names, types (string, number, array, object), and a brief description of each field. Example: 'Returns {rsi: number, ema_20: number, ema_50: number, ema_200: number, atr: number, bollinger_bands: {upper: number, lower: number}, vwap: number, z_score: number, trend: string, confluence: {alignment_score: number}, derivatives: {funding_rate: number, open_interest: number, spread: number}, market_regime: string}'.
• Replace 'timeframes' string parameter with enum constraint: ['5m', '15m', '1h', '4h', '1d', '1w']. Add validation in server code to reject invalid timeframes before API call. Update description to list valid options explicitly.
• Add explicit recovery guidance to tool descriptions for paid endpoints. Example suffix: 'If payment is required (402), the response will include payment terms (price_usdc, network, pay_to). Set CEREBRUS_WALLET_KEY and install payment extras to auto-pay, or pay manually and retry.' This surfaces the fallback behavior to LLM context.
• Enhance 'lookback_hours' parameter description: 'Number of hours of historical funding rate data to analyze (1-168, default 24). Used to compute min/max/average funding rates over the period. Longer lookback provides trend context; shorter captures recent extremes.'
• Add a limit parameter to cerebrus_list_coins with default 30 and max 100. Document that results are paginated and large result sets are truncated. If the API does not support pagination, cap server-side result to 30 and document this in the tool description.