MCP server for interacting with Binance cryptocurrency exchange APIs, including spot trading, futures, and arbitrage strategies
Server has 9 tools with basic schemas and descriptions, but significant quality gaps. All tools have descriptions (10-50 chars, mostly under 100), which is below the 194-char baseline. Input schemas are present with type declarations, but lack depth: no enums for constrained inputs (e.g., 'side' should be enum BUY|SELL), no min/max bounds on numeric params, no format specs. Parameter descriptions are minimal (5-15 chars) vs. 72-char baseline. Output schemas are undocumented, LLMs cannot predict response structure. Error handling is generic ('error': response.text) with no recovery guidance. Critical security issue: API keys passed as command-line args, not server-side injection. Tool composition is reasonable (single responsibility), but descriptions lack WHEN/WHY context. Naming is verb-first and clear (get_*, place_*, execute_*), which is good.
Cancel a specific order.
Execute hedge arbitrage based on funding rate.
Find arbitrage pairs based on funding rate, volume, and rate direction stability.
Get the balance of a specific cryptocurrency asset.
Get funding rate history.
Get open orders for a symbol.
Get the current price of a cryptocurrency pair.
API credentials exposed as command-line arguments (--binance-api-key, --binance-secret-key). Credentials must never appear in tool parameters or startup args; use server-side secret injection via environment variables or vault. Agent traces log all parameters, secrets leak into logs and prompt history.
Parameter descriptions are minimal (5-15 chars) vs. 72-char baseline. E.g., 'side' param has description 'BUY or SELL.' but lacks context on when to use each. 'quantity' lacks format/range guidance (is it a decimal? integer? min/max?). LLMs cannot infer constraints from names alone.
No enum constraints for constrained inputs. 'side' parameter in place_market_order should declare enum: ['BUY', 'SELL']. Free-form strings invite hallucinated values. Enums are self-documenting and let LLMs pick valid options without guessing.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
Get recent trade history for a pair.
Place a market order to buy or sell.
Output schemas are undocumented. LLMs cannot predict response structure (e.g., does get_symbol_price return {price, symbol} or {data: {price, symbol}}?). Tool descriptions must document return types and field names so agents can chain calls and extract data reliably.
Error handling is generic ('error': response.text). No recovery guidance. E.g., if place_market_order fails with 'Insufficient balance', the LLM has no hint to call get_account_balance first. Errors must tell the agent what to do next.