Static source inference · medium confidence · detected: Logging
Deprecated protocol patterns detected
Summary
This server presents 8 tools with explicit schemas and descriptions visible in the source code. Naming follows verb_noun conventions (get_*, set_, clear_), which is good. However, there are significant gaps in parameter descriptions, output schema documentation, and error handling guidance. Most tools have shallow descriptions (under 100 chars) that lack context about WHEN to use them or WHAT they return. Parameters like 'symbol' in get_market_summary have defaults but lack validation constraints. The server uses STDIO transport only, which limits remote accessibility. Overall, this falls into the 'Fair to Poor' range, definitions exist but lack production-grade rigor.
Tools (8)
clear_bufferwritesource verified73/100
Clear the market data message buffer.
get_alertsread onlysource verified77/100
Get all active and triggered alerts with their current status.
get_latest_messageread onlysource verified80/100
Get the most recent market message received from the WebSocket stream.
get_market_summaryread onlysource verified75/100
Get structured order book analysis (best bid/ask, spread, volumes, imbalance, price change).
get_recent_messagesread onlysource verified80/100
Get the last N market messages from the stream buffer (max 500).
get_stream_statusread onlysource verified77/100
Check the status of the WebSocket connection and how many messages have been buffered.
Output schemas not documented for any tool. LLMs cannot predict what fields get_latest_message, get_market_summary, or get_alerts will return. This violates pattern:tool and forces the agent to parse free-form text, risking errors.
Parameter descriptions lack validation constraints. 'symbol' in get_market_summary defaults to 'BTC-USD' but does not specify format (e.g., '[A-Z]{1,5}-[A-Z]{3}') or valid range. 'threshold_pct' in set_percent_change_alert lacks min/max (e.g., 0.01-100). LLMs will pass invalid values without guidance on format.
Destructive operations (clear_buffer, set_price_alert, set_percent_change_alert) lack confirmation or dry-run support. The description for clear_buffer does not warn that it is irreversible. Agents may clear the buffer unintentionally without recovery path.
Recommendations
Document output schemas for every tool. For get_latest_message, specify: {type: 'object', properties: {message: {type: 'string'}, timestamp: {type: 'string'}, symbol: {type: 'string'}, ...}, required: [...]}. This enables LLMs to chain tools and extract the right fields.
Expand tool descriptions to 50-200 chars, answering: What does it do? When should the LLM call it? What does it return? Example for get_latest_message: 'Return the most recent market data message from the WebSocket stream. Call this first to check if the stream is active. Returns: raw message string, timestamp, and symbol. Returns null if no messages received yet.'
Add format and range constraints to parameters. For 'symbol', add description: 'Market symbol in format XXX-YYY (e.g., BTC-USD, ETH-USD). Case-sensitive.' For 'threshold_pct', add: 'Absolute percentage threshold (0.01 - 100, e.g., 2.5 for ±2.5% move).'
Document failure modes and recovery paths in descriptions. For get_latest_message: '...If the buffer is empty, call get_stream_status to verify the WebSocket is connected. If not, the server may need to reconnect.'
Add idempotentHint and destructiveHint tool annotations (MCP 2026-07-28 feature). Mark clear_buffer with {destructiveHint: true} and set_price_alert with {idempotentHint: false} to warn agents about side effects.
Clarify alert semantics. Add description: 'set_price_alert and set_percent_change_alert create different alert types that coexist. Call get_alerts to see all active and triggered alerts. Triggered alerts remain in the response until manually cleared.'
Spec posture evidence
Inferred effective spec: 2026-07-28+.
Relies on Logging (deprecated) - log to stderr or use OpenTelemetry
Score history
Overall score trend
↑ 13 points across a rubric change (v1 → v2)
57/100
Scored
Grade
Overall
Spec posture
Rubric
2026-09-22
D
57
2026-07-28+
v2
2026-03-09
F
44
-
v1
set_price_alertwritesource verified75/100
Set a price alert that triggers when a symbol's price crosses a threshold.
Tool descriptions are generic and lack context about WHEN to use them. E.g., 'Get the most recent market message' does not explain: What is a 'message'? When would I call this vs get_recent_messages? What fields does it contain? Does it fail if no messages exist? Descriptions should be 50-200 chars and answer these questions.
No error handling guidance. Tools do not document what happens on failure (e.g., WebSocket disconnected, buffer empty, invalid symbol). Error responses from call_tool return plain text but do not guide the LLM to recovery (e.g., 'No messages received yet' does not suggest calling get_stream_status or waiting).
Tool interdependencies not documented. set_price_alert and set_percent_change_alert both create alerts, but descriptions do not explain how they differ, when to use each, or if they can coexist. get_alerts returns 'active and triggered' but does not describe the structure or how to act on triggered alerts.
set_price_alertset_percent_change_alertget_alerts
Implement per-item error details. If set_price_alert fails (e.g., invalid symbol), return: {success: false, error: 'Invalid symbol: got BTC, must be in format XXX-YYY (e.g., BTC-USD)'}.
Add pagination/limits to get_recent_messages and get_alerts. Document max page size and total count in response: 'Returns last N messages (max 500). Response includes: messages (array), total_buffered (int), oldest_message_id (int or string for pagination).'
Define the message structure. Document what fields are in a 'market message'. E.g.: 'Each message is a JSON object: {timestamp: ISO8601, symbol: string, bid: number, ask: number, volume: number, ...}. See schema for full structure.'
Add WRITE risk context to descriptions. For clear_buffer: 'Clears the entire message buffer. This is irreversible, recent market data will be lost. Use with caution.' For set_price_alert: 'Creates a new alert. This alert persists until manually removed via get_alerts and (implicitly) deletion. Calling this multiple times with the same symbol may create duplicate alerts unless overwrite is supported (clarify).'