An MCP server exposing trading analytics (indicators, portfolio, risk, backtests) to LLM agents, with read/write separation and human-in-the-loop guardrails.
TradeMCP demonstrates solid fundamentals: all 5 tools have clear verb-noun naming (trade_get_*, trade_list_*), comprehensive descriptions (150-250 chars each), and complete input schemas with types and descriptions. Tool annotations (readOnlyHint, destructiveHint, idempotentHint) are properly declared. However, output schemas are documented only in docstrings, not in the tool registration itself, LLMs cannot introspect them. Error handling returns formatted text but lacks recovery guidance (e.g., 'Try X next'). Parameter descriptions are good but could be more prescriptive about constraints. No batch operations or composition patterns evident.
Get full metrics for a single backtest run by its ID. Use trade_list_backtests first to find a valid run_id, then call this for the detailed performance breakdown (returns, Sharpe/Sortino, drawdown, win rate, profit factor, capital curve endpoints).
Get the latest technical indicators for a single NSE symbol. Returns a snapshot of common momentum/trend indicators (RSI, EMAs, MACD, ATR) plus a derived bullish/bearish signal. Read-only; computes nothing on live capital.
Get the current portfolio: holdings, cash, equity, and unrealized P&L. Read-only snapshot of all open positions with per-position and aggregate profit/loss.
Get aggregate risk metrics for the current portfolio. Surfaces concentration, beta, Value-at-Risk, Sharpe, drawdown, risk-budget usage, and any open circuit breakers. Read-only.
List available backtest runs, most useful for discovering run IDs. Returns a paginated list of run summaries. Pair this with trade_get_backtest to drill into a specific run's full metrics.
Output schemas documented only in docstrings, not in tool registration. LLMs cannot introspect return types from the MCP protocol layer.
Error handling returns format_error(exc) text but lacks actionable recovery guidance. Errors should suggest next steps (e.g., 'Symbol not found. Try trade_list_backtests() to discover valid symbols.').
Parameter 'symbol' in trade_get_indicators lacks format constraints. Description says 'case-insensitive' but no regex pattern or enum. Should specify valid NSE ticker format (uppercase letters, 1-10 chars typical).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
trade_list_backtests accepts limit (1-50) and offset but does not return total_count or next_cursor. Pagination metadata is missing, forcing agents to guess when results are exhausted.
No batch operations. If an agent needs indicators for 10 symbols, it must call trade_get_indicators 10 times sequentially. A trade_get_indicators_batch(symbols: [str]) would reduce latency and token waste.