Loading ToolBench
A quantitative trading research and execution platform with data collection, backtesting, and strategy management capabilities. Not an MCP server.
QuantWorkstation exposes 9 tools, all named 'collect' or 'collect_mer', creating severe naming ambiguity. No tool descriptions are visible in the source code provided, only brief inline docstrings in the Makefile and pyproject.toml. Input schemas are partially present (symbol, timeframe, url, series parameters documented) but lack formal JSON Schema definitions in the code. All tools are marked WRITE risk but lack error handling guidance, recovery instructions, or idempotency guarantees. No output schemas documented. The server appears to be a data collection CLI wrapper, not a production MCP server with proper tool registration.
Fetch earnings surprises for key constituents and write to ArcticDB.
Fetch EIA petroleum stock series and write to ArcticDB macro library.
Fetch and store bars for one symbol. Incremental if data exists.
Download Baker Hughes NA rig count and write to ArcticDB macro library. Incremental: reads the last stored date for each series and appends only rows newer than that date. Safe to run repeatedly — no duplicates.
Fetch current BDTI and append to ArcticDB macro library. Skips the write if today's date is already stored.
Download CFTC COT data and write to ArcticDB cot library.
Fetch economic calendar and write to ArcticDB calendar library. Uses a rolling window of -7 days to +30 days relative to today. Idempotent: write_series deduplicates by index before writing.
Seven tools named 'collect' with no distinguishing prefix or suffix. LLMs cannot disambiguate which tool to call for a specific data source (crypto vs macro vs earnings). Violates verb_noun naming convention and creates unresolvable tool selection ambiguity.
No formal input schemas visible in source code. Parameter descriptions exist in docstrings but are not registered as JSON Schema with type definitions. Cannot verify schema completeness or validate LLM-generated inputs.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 27 | 2026-07-28+ | v2 |
Collect all configured crypto symbols.
Fetch EIA MER degree-day series and write to ArcticDB macro library.
No output schemas documented. LLMs cannot predict what data these tools return, making it impossible to chain them with downstream tools or extract specific fields.
All tools marked WRITE risk but no error handling, recovery guidance, or idempotency guarantees documented. Agents cannot determine if retrying a failed collect is safe or if it will create duplicates.
Transport mechanism unknown. Makefile and pyproject.toml show CLI scaffolding (qw command) but no MCP server registration, HTTP endpoint, or STDIO handler visible. Cannot confirm this is a functional MCP server.