Loading ToolBench
A WebMCP server for a pet product store that provides AI agents with tools to check device compatibility against household profiles, search products, and add compatible items to cart. Focuses on preventing returns due to mismatches (Wi-Fi bands, LTE coverage, regional app availability, pet weight limits, power requirements).
MintPaw WebMCP exhibits significant quality gaps across naming, descriptions, and schema consistency. Tool names lack clear verb-noun patterns (e.g., 'set_household_profile' vs 'update_household_profile'; 'find_compatible_alternative' is vague). Descriptions vary wildly in quality: some are detailed (set_household_profile: 156 chars), others are generic or domain-specific without context (list_orders, add_order use Chinese status enums without explanation). Parameter descriptions are sparse or missing entirely (e.g., 'category' and 'search' in find_compatible_alternative lack guidance on format/constraints). Output schemas are not documented anywhere in the source. The codebase mixes three distinct domains (pet product compatibility, restaurant order management, news search) without clear separation, suggesting tool sprawl. Error handling is absent from visible code, no recovery guidance, no validation messages. Schemas are present but incomplete: many parameters lack type constraints, and no enum validation is shown for critical fields like 'stage' in list_orders.
Add a product that is compatible with the current household profile to the shopping cart. Only available when a household profile is set. Verifies compatibility before adding.
Create a new order and place it in 'waiting for order' status. Use when the user says 'two servings of XX' or 'place an order'.
Advance a specific order to the next status (待接单 → 备餐中 → 已出餐). Only available when there are pending orders.
Check compatibility of all products in the catalogue against the current household profile. Returns a list of products with verdicts (ok/caution/blocked/unknown) across five axes: network, region, power, pet, extras. Refuses to run without a profile set.
Switch the order board to a specific status view. Pass 'all' to show all orders.
Search the catalogue for products compatible with the current household profile. Returns products that pass the compatibility check, optionally filtered by category or search term.
Output schemas completely undocumented. No tool describes what fields it returns, forcing LLMs to guess structure and plan downstream calls blindly.
Domain confusion: tools span pet product compatibility (set_household_profile, check_compatibility), restaurant order management (list_orders, add_order, filter_orders, advance_order), and news search (search_news). No clear separation or context for when each domain applies.
Parameter descriptions missing or vague. 'category' and 'search' in find_compatible_alternative lack format guidance. 'stage' enum values in list_orders/filter_orders are Chinese strings (待接单/备餐中/已出餐) with no English explanation or context.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |
List out current all orders and their status (待接单/备餐中/已出餐). Used to answer questions like 'how many orders now' and 'which ones are not done yet'.
Search news articles on this site by keyword. Can specify sort order and result limit. Use to answer questions like 'latest news about X'.
Set or update the household profile (pet species/count/weight, home Wi-Fi bands, country, cellular plan status). This profile gates the availability of compatibility-checking tools and is the foundation for all verdicts.
Naming ambiguity: 'find_compatible_alternative' is vague (alternative to what?). 'filter_orders' is a write operation (state change) but named like a read-only filter. 'advance_order' lacks clarity on direction (forward/backward?).
No error handling guidance. No tool describes what happens on failure, what to retry, or how to recover. E.g., add_compatible_to_cart silently fails if profile not set; no error message guides the LLM.