An MCP server that gives AI agents live market prices and candles. Data only, no trading advice.
Four tools with clear verb-noun naming (get_price, get_prices, list_symbols, get_candles). All have descriptions and input schemas with types. However, descriptions lack depth on WHEN to use each tool and lack dependency hints. Parameter descriptions are present but minimal (e.g., 'the coin ticker' without format constraints). Output schemas are not documented, LLMs cannot infer what fields to expect. No error handling guidance (e.g., what happens if symbol is invalid). Missing enum constraints on 'interval' parameter despite fixed valid values. No pagination support despite potential for large result sets.
Get recent OHLCV candles (price history) for a asset. Args: symbol: the coin ticker, e.g. "BTC". interval: one of 1m, 5m, 15m, 1h, 4h, 1d. count: how many recent candles to return (default 100). Returns candles oldest-first, each with open/high/low/close/volume.
Get the current price of a asset. Args: symbol: the coin ticker, e.g. "BTC", "ETH", "SOL". Returns the symbol, its current price (USD), and an as-of timestamp (ms).
Get current prices for several assets at once. Args: symbols: list of tickers, e.g. ["BTC", "ETH", "SOL"].
List all symbols this server can price.
Output schemas not documented. LLMs cannot infer return structure (e.g., get_price returns {symbol, price, as_of} but this is not declared in tool definition). Forces LLMs to guess field names and types.
Missing enum constraint on 'interval' parameter in get_candles. Description says 'one of 1m, 5m, 15m, 1h, 4h, 1d' but no formal enum in schema. LLMs may hallucinate invalid intervals like '2h' or '30m'.
No error handling guidance. If symbol is invalid or API fails, LLMs receive raw exceptions (ValueError, urllib errors) with no recovery hints. Should return 'Symbol not found. Call list_symbols() to see available coins.'
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2026-07-28+ | v2 |
No pagination or result limits documented. get_prices accepts unbounded list of symbols; get_candles can return up to 100+ candles. Large responses waste tokens. Should cap results and document limits in descriptions.
Parameter descriptions lack format constraints. 'symbol' says 'e.g. BTC' but doesn't state case-insensitivity, length limits, or valid character set. 'count' lacks min/max bounds (1 - 1000?).