Model Context Protocol server for Coinbase public API integration with cryptocurrency price data, market statistics, asset search, and demo wallet functionality
The server defines 15 tools with basic naming and input schemas, but suffers from significant gaps in description quality, output schema documentation, and error handling guidance. Most tool descriptions are terse (10-40 chars), many parameters lack descriptions or type clarity, and output structures are not documented. Tools like 'calculate_beer_cost' and 'buy_virtual_beer' are novelty/demo features that dilute focus. No tool annotations (readOnlyHint, destructiveHint) present despite clear risk classifications (READ_ONLY vs WRITE). Error handling is absent from visible code, no recovery guidance, no input validation messages. The API server (api-server/src/index.ts excerpt) shows only skeleton; full schema validation and error construction are not visible in provided source. Rating reflects: strong naming convention (verb-noun), present but minimal descriptions, schemas visible but output undocumented, zero error guidance, no permission checks or secret injection patterns evident.
Perform technical analysis on price data
Buy virtual beer with cryptocurrency from the demo wallet
Calculate how much cryptocurrency equals the cost of beer
Get detailed information about a specific asset
Get latest cryptocurrency news and market sentiment
Get exchange rates for a base currency
Get historical price data for a cryptocurrency pair
Output schemas are not documented. Tools like get_spot_price, get_historical_prices, get_wallet_balance declare what they do but do NOT describe their response structure. LLMs cannot plan downstream calls or extract fields without knowing what data is returned.
Tool descriptions are too short (10-40 chars) to guide LLM selection. E.g., 'Get current spot price for a cryptocurrency pair' (49 chars) lacks context on WHEN to use it vs analyze_price_data or get_market_stats. Production baseline: 194 chars average.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 44 | - | v1 |
Get 24-hour market statistics for a currency pair
Get list of popular cryptocurrency trading pairs
Get the current spot price for a cryptocurrency pair
Get transaction history from the demo wallet
Get current demo wallet balance
Reset the demo wallet to initial state
Search for cryptocurrency and fiat assets
Simulate purchasing cryptocurrency with fiat currency
Parameters lack descriptions. E.g., get_popular_pairs has NO parameters listed, but likely accepts a limit or category parameter internally. get_wallet_balance similarly has no visible parameters. Undescribed params force LLMs to guess.
No tool annotations despite clear risk classifications. Tools marked WRITE (simulate_purchase, buy_virtual_beer, reset_wallet) should declare destructiveHint=true. READ_ONLY tools should declare readOnlyHint=true. Annotations guide agent safety and planning.
No error handling guidance visible. Tools do not document what happens on invalid input (e.g., invalid currency pair), missing asset, or rate limiting. Agents have no recovery path.
Novelty/demo tools dilute focus. 'calculate_beer_cost' and 'buy_virtual_beer' are gimmicks that reduce signal-to-noise ratio. Remove them or clearly mark as demo-only, separate from production crypto tools.
No pagination guidance. Tools like search_assets and get_crypto_news accept a 'limit' parameter but do not document total count, next_cursor, or whether results are truncated. Large result sets risk context window overflow.
Parameter constraints not documented as enums/patterns. E.g., analyze_price_data has period enum (['1d','7d','30d','1y']) and metrics enum, these are good, but get_historical_prices 'period' has enum ['hour','day'] yet description does not explain what 'hour' and 'day' mean or whether they are only valid combinations.