Loading ToolBench
Exposes the DataSignals Lab suite of scored financial, regulatory, biotech and market-data products as MCP tools that an AI agent can call directly. Each tool is a thin proxy that runs the corresponding published Apify Actor and returns its scored results.
14 tools with consistent structure but significant quality gaps. All tools have descriptions (50-80 chars, below the 194-char baseline) and input schemas with typed parameters. However, descriptions are generic and lack actionable context for LLM selection. No output schemas documented. Error handling returns structured errors but lacks recovery guidance. Tool names lack action verbs (insider_buying, app_reviews are nouns, not verb_noun pattern). Parameters accept arrays but lack enum constraints or format specifications. No tool annotations (readOnlyHint, idempotentHint). The 'usage' tool is a discovery/reporting tool but named generically. Overall: functional but below production baseline for agent-optimized tools.
Run the activist_13dg DataSignals Actor and return its scored dataset items.
Run the app_reviews DataSignals Actor and return its scored dataset items.
Run the biotech_catalysts DataSignals Actor and return its scored dataset items.
Run the congress_trades DataSignals Actor and return its scored dataset items.
Run the crypto_momentum DataSignals Actor and return its scored dataset items.
Run the fda_actions DataSignals Actor and return its scored dataset items.
Tool names lack action verbs. All 13 data-query tools use noun patterns (insider_buying, app_reviews) instead of verb_noun (query_insider_buying, search_app_reviews). LLMs infer intent from names, noun-only names force description parsing.
Descriptions are generic and lack LLM-optimized context. All 13 query tools use identical template: 'Run the [X] DataSignals Actor and return its scored dataset items.' No explanation of WHEN to use each tool, what domain problem it solves, or how it differs from similar tools. Baseline is 194 chars; these are ~70 chars.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Run the formd_funding DataSignals Actor and return its scored dataset items.
Run the gov_contracts DataSignals Actor and return its scored dataset items.
Run the hedge_fund_13f DataSignals Actor and return its scored dataset items.
Run the insider_buying DataSignals Actor and return its scored dataset items.
Run the insider_selling DataSignals Actor and return its scored dataset items.
Run the nih_funding DataSignals Actor and return its scored dataset items.
Run the sec_8k DataSignals Actor and return its scored dataset items.
Return the current usage ledger for the caller, showing billable results pulled through this server and estimated cost.
No output schemas documented. Tools return structured dicts with 'ok', 'count', 'items', 'billed', 'billed_usd', 'summary' fields, but LLMs have no schema to parse them. Baseline: 100% of A+ tools document return types. This blocks downstream tool chaining and forces LLMs to infer structure.
Parameters lack constraints and format guidance. 'tickers', 'apps', 'funds', 'companies', 'assets' accept arrays but have no enum, pattern, or length constraints. No guidance on valid ticker formats (NYSE vs NASDAQ), app ID formats, or company identifier schemes. Invites hallucinated invalid values.
No tool annotations. Tools lack readOnlyHint (all are read-only), idempotentHint (all are idempotent), or destructiveHint (none are destructive). Annotations help agents reason about retry safety and side effects. Current spec (2026-07-28) supports these; omission is a missed signal.