Loading ToolBench
A Model Context Protocol server providing sports intelligence and fan token analysis tools for multiple sports including football, cricket, MMA, and Formula 1. Integrates SportMind domain knowledge with LLM reasoning for sports prediction and signal generation.
SportMind MCP server has 5 tools with reasonable naming (verb_noun pattern) and detailed descriptions (150-250 chars each). However, critical gaps undermine quality: input schemas are present but lack proper JSON Schema type definitions for most parameters; parameter descriptions are minimal or missing; output schemas are completely undocumented; no error handling guidance; no tool annotations (readOnlyHint). The benchmark script shows domain sophistication, but the MCP tool definitions themselves are incomplete for production use. Average per-tool score: 42/100.
Check the macro state first — the macro modifier gates every fan token signal. Returns crypto cycle phase and macro_modifier. If modifier is below 0.75, the entire crypto market is in a bear phase and sporting signals lose reliability.
Execute the full six-step SportMind reasoning chain for an event. Steps: 1) Macro check, 2) Competition classification, 3) Athlete availability, 4) Signal computation, 5) DeFi/liquidity check, 6) Confidence output.
Generate a pre-match signal using SportMind intelligence. Takes sport, event details, and optional use_case. Returns SportMind Score (SMS), confidence tier, recommended action (ENTER/WAIT/ABSTAIN), and key modifiers applied.
Sport-specific analysis with routing to correct reasoning model. Handles football (competition tier + lineup window), cricket (format-first rule), MMA (weigh-in protocol), Formula 1 (qualifying delta), and basketball.
Load the SportMind intelligence stack for a given sport and use case. Returns list of skill files loaded in correct order: macro → market → sport domain → athlete → fan token.
Output schemas completely undocumented. No tool returns a documented structure, LLMs cannot plan downstream calls or extract specific fields. sportmind_signal claims to return 'SMS, confidence tier, recommended action' but no schema is visible.
Input parameters lack type definitions in JSON Schema. sportmind_signal lists 'sport', 'event_id', 'home_team', etc. as strings but provides no enum constraints, length limits, or format specifications. LLMs cannot validate inputs before calling.
No error handling guidance. Tools declare READ_ONLY risk but provide no recovery instructions. If sportmind_macro fails (e.g., crypto data unavailable), the LLM has no guidance on retry strategy or fallback.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are marked READ_ONLY in metadata but this is not reflected in MCP tool definitions. Agents cannot determine idempotency or retry safety from the schema.
Parameter descriptions are minimal. 'sport' is described as 'Sport type (football, cricket, ...)' but lacks guidance on when to use each. 'use_case' in sportmind_signal has no description of valid values or selection criteria.