Loading ToolBench
Paid MCP server: real-time local business intelligence for AI agents. Exposes business search, website auditing, lead scoring, and tech stack detection.
BizIntel MCP has 8 well-named tools with clear action verbs (audit_, search_, get_, find_, score_). All tools have descriptions (avg ~120 chars, within baseline 34-392 range). Input schemas are present with type declarations and parameter descriptions. However, output schemas are NOT documented, responses are inferred from code but not formally declared in tool definitions. Error handling is minimal; most tools return generic dicts without recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being read-only. Parameter descriptions are adequate but lack format constraints (e.g., 'limit' has no min/max bounds). Composition is strong, tools chain well (search → audit → score), and responses include IDs needed for downstream calls.
Audit a single website URL across SSL, mobile, speed, contact, and booking signals. Returns a score 0-100 and detailed breakdown.
Audit multiple website URLs concurrently. Max 20 URLs per call. Results sorted by score (worst first = best leads first).
Find high-potential leads in a niche/location that have websites but lack online booking systems. Audits websites and filters for booking gap opportunity.
Find high-potential leads in a niche/location that lack a website. Searches businesses and filters those without websites, scoring by demand signal.
Retrieve detailed information for a single business by name and city. Primary: Yelp /matches endpoint. Fallback: OpenStreetMap Nominatim.
Output schemas not formally documented. Tool definitions show input schemas but no declared return types. LLMs cannot plan downstream calls or extract fields without inferring structure from descriptions.
Numeric parameters lack bounds. 'limit' in search_businesses and find_* tools has no min/max declared. 'limit' defaults to 20 but max is 50, this constraint is only in description text, not in schema.
No tool annotations despite all 8 tools being read-only. Missing readOnlyHint in tool definitions. Agents cannot distinguish safe tools from destructive ones without explicit hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2026-07-28+ | v2 |
Lightweight tech stack lookup for a website URL. Uses the audit cache to avoid duplicate fetches. Returns detected CMS, booking system, email provider, analytics, and server info.
Composite lead scoring combining audit signal and business signal. Scores 0-100 with breakdown of scoring factors.
Search for businesses by niche and location. Primary: Yelp Fusion API. Fallback: OpenStreetMap Overpass.
Error handling lacks recovery guidance. Tools return generic error dicts (e.g., 'not found', 'no urls provided') without suggesting next steps. No categorization of retryable vs fatal errors.
Parameter 'state' in search_businesses and find_* tools is optional but no guidance on when it's required vs optional. Description says '(optional)' but does not explain behavior when omitted.