AI-powered options research, strategy, risk analysis and paper-trading platform.
This server has critical definition gaps across nearly all tools. Most tool descriptions are present but lack actionable detail for LLM selection. Input schemas are partially visible but many parameters lack type definitions or descriptions. No output schemas are documented. Tool names follow verb_noun convention (positive), but parameter naming is inconsistent. Error handling is minimal, no recovery guidance, no categorization of retryable vs fatal errors. The `track_pnl_and_monitor_positions` tool combines multiple concerns (tracking AND monitoring AND triggering actions), violating single-responsibility. No tool annotations (readOnlyHint/destructiveHint) despite having both read-only and irreversible operations. Security: no evidence of secret injection for API keys (OpenRouter key appears in code). Overall, this reads like a prototype rather than production-grade tooling.
Approve or reject a trade candidate based on risk engine rules: max loss validation, reward-risk ratio (minimum 1.5), per-trade risk limit (2% default), and total portfolio risk limit (10% default)
Evaluate a proposed trade candidate using AI analysis of risk-reward ratios, maximum loss, maximum profit, breakeven points, and option leg structure
Execute a trade candidate by submitting single-leg or multi-leg option orders to Alpaca after risk approval
Fetch historical bars and use LLM to perform advanced ICT (Inner Circle Trader) SMC analysis on multi-timeframe price action, identifying Daily Bias, FVGs, Order Blocks, and Liquidity Sweeps
Fetch general market news and apply Bridgewater Associates macro framework to determine current macro regime and equity outlook
Tool `track_pnl_and_monitor_positions` violates single-responsibility principle: it tracks P&L, monitors positions, AND triggers take-profit/stop-loss actions. Should split into separate tools: `get_portfolio_pnl`, `monitor_positions`, `trigger_exit_signals`.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. E.g., `get_option_chain` returns option data but structure is undocumented; `evaluate_candidate` returns AgentDecision but this is not visible in tool definition.
Parameter descriptions are missing or generic. E.g., `candidate` parameter in `evaluate_candidate` and `execute_trade` is typed as 'object' with minimal guidance on required fields. `market_context` in `evaluate_candidate` lacks format/length constraints.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | <=2025-11-25 | v2 |
Retrieve option chain data for a given symbol with optional filtering by option type and expiration
Fetch latest news for a symbol and use AI to gauge sentiment (BULLISH, BEARISH, or NEUTRAL) to establish directional edge
Health check endpoint that returns service status, name, and version
Track portfolio P&L and monitor open positions with automatic take profit (50% gain) and stop loss (30% loss) triggers
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). `execute_trade` is irreversible but unmarked. `health_check` and read-only tools lack readOnlyHint. Agents cannot distinguish safe from dangerous operations.
Error handling is absent. No recovery guidance, no error categorization (retryable vs fatal), no actionable error messages. E.g., `evaluate_candidate` returns generic 'Error connecting to OpenRouter' with no guidance on retry or fallback.