Loading ToolBench
Market Intelligence MCP Server for AI agents. Serves regime-aware market intelligence, Redis-backed candle reads, cross-exchange spread detection, signal stream access, and graph relationships (leader-lag, correlations, sector rotation).
MIA server has 21 well-named tools with clear action verbs (get_*, all READ_ONLY). Descriptions are present and substantive (100-250 chars typical), explaining what each tool returns and when to use it. Input schemas are visible with typed parameters and defaults. However, critical gaps exist: (1) No output schemas documented, LLMs cannot predict response structure for downstream chaining; (2) Parameter descriptions lack constraints (e.g., 'symbol' accepts any string, no format hint; 'timeframe' has no enum of valid values like '1h','4h','1d'); (3) No error handling guidance, tools return data or fail silently with no recovery hints; (4) No tool annotations (readOnlyHint, idempotentHint) despite all being read-only; (5) Composition risk: 21 tools invite decision paralysis, get_market_intelligence aggregates regime+signals+confluence, but agents may still call individual tools redundantly. Naming is strong (verb_noun, clear distinctions), descriptions are above-average length, but lack LLM-actionable constraints and output documentation.
Get active trading signals for a symbol
Get list of actively monitored symbols
Get OHLCV candles for a symbol with Redis-first caching and exchange fallback
Get capital flow analysis across markets
Get confluence score indicating how many indicators align on a signal
Get correlation matrix between symbols
Get price spread for a symbol across exchanges
No output schemas documented. LLMs cannot predict response structure (fields, types, nesting) for downstream tool chaining. E.g., get_market_intelligence returns 'regime_data', 'node_data', 'confluence_score', 'bias', 'key_levels', but no schema tells the agent what fields exist or how to extract them for the next call.
Parameter constraints missing. 'symbol' accepts any string (no format hint like 'BTC/USDT' pattern); 'timeframe' has no enum of valid values ('1h','4h','1d','1w'); 'exchange' defaults to 'aggregated' but no enum of valid exchanges. LLMs will hallucinate invalid values.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 50 | <=2025-11-25 | v2 |
Get lead-lag relationships between symbols
Full market intelligence snapshot for a symbol. Returns regime, bias, confluence score, key levels, active signals, and session context in a single call. This is the primary tool for agents making trading decisions — it aggregates all intelligence layers rather than requiring multiple separate calls.
Get momentum indicators (RSI, Stochastic, Williams %R, ROC, Momentum) for a symbol
Get candles across multiple timeframes in a single call
Get order book snapshot for a symbol
Get market regime for a symbol (trending up/down, ranging, etc.)
Get session context and weighting for current trading session
Get support and resistance levels for a symbol
Get system health status including Redis, adapters, and data freshness
Get current ticker data for a symbol
Get top alpha-generating markets
Get trend indicators (MACD, ADX, Parabolic SAR) for a symbol
Get volatility indicators (Bollinger Bands, ATR, Keltner Channel, Donchian Channel) for a symbol
Get volume indicators (OBV, MFI, VWAP) for a symbol
No error handling guidance. Tools return data or fail silently with no recovery hints. E.g., if 'symbol' is invalid or not found, what does the agent do? No error message tells it to call get_active_symbols() first or suggests valid symbols.
No tool annotations. All 21 tools are read-only (no side effects), but readOnlyHint is not set. Agents cannot distinguish safe tools from destructive ones without explicit hints.
Tool composition risk: 21 tools invite decision paralysis. get_market_intelligence aggregates regime+signals+confluence, but agents may still call individual tools redundantly. No guidance on which tool to call first or when to prefer the aggregated tool.