A collection of three MCP servers providing access to pharmaceutical data: products, companies, and customer comments/reviews from the Unofficial Farma database
Strong foundation with clear naming conventions (verb_noun pattern consistently applied across all 9 tools) and comprehensive parameter descriptions. All tools have well-structured input schemas with type definitions and parameter-level documentation. However, tool descriptions, while present, are verbose (200-300+ chars) and could be more concise for LLM parsing efficiency. Output schemas are not explicitly documented in the source code, only inferred from Pydantic model validation. No error handling guidance visible; tools appear to return raw API responses without recovery hints. All tools are read-only (no destructive operations), reducing security complexity but also limiting domain coverage.
Check that the comment service is reachable and answering requests. Use this to tell a service outage apart from a search that simply found nothing.
Check that the company service is reachable and answering requests. Use this to tell a service outage apart from a search that simply found nothing.
Check that the product service is reachable and answering requests. Use this to tell a service outage apart from a search that simply found nothing.
List customer reviews left on medicines in Unofficial Farma. Returns comments in insertion order, across every product and without any filtering, so read product_id on each item to know which medicine it refers to. Use search_comments instead to read the reviews of one medicine.
List the pharmaceutical companies registered in Unofficial Farma. Returns companies in insertion order, without any filtering. Use this to browse the whole catalogue or to page through it; use search_companies instead when looking for a company by name.
Output schemas not documented in tool definitions. Pydantic models (Comment, Company, Product) are validated internally but not exposed in MCP tool schemas. LLMs cannot predict response structure without explicit schema documentation.
Tool descriptions exceed 200 characters and contain implementation details ('Returns comments in insertion order, across every product and without any filtering, so read product_id on each item...') that distract from the core action. LLM-optimized descriptions should be 50-150 chars focusing on WHAT and WHEN.
No error handling or recovery guidance visible. Tools appear to return raw API responses without categorizing errors as retryable, user-fixable, or fatal. No guidance on what to do if a service is unreachable or returns empty results.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2026-07-28+ | v2 |
List the medicines registered in Unofficial Farma. Returns products in insertion order, without any filtering. Use this to browse the whole catalogue or to page through it; use search_products instead when looking for medicines by name, maker, dosage or price.
Search customer reviews by medicine, author or wording. Filters combine with AND, so passing several narrows the result. The text filters (author_name, author_email, content) are case-insensitive partial matches, so "dor" finds every review mentioning it. product_id is an exact match and is the usual way to read the reviews of one medicine. Omitting every filter returns all comments, exactly like list_comments.
Search pharmaceutical companies by name. The name filter is a case-insensitive partial match, so "vale" matches "Laboratório Vale Verde". Omitting it returns every company, exactly like list_companies.
Search medicines by maker, name, description, price or dosage. Filters combine with AND, so passing several narrows the result. The text filters (name, description, dosage) are case-insensitive partial matches, so "ibup" matches "Ibuprofeno". company_id and price are exact matches: price will not find a range, only medicines costing that precise amount. Omitting every filter returns all products, exactly like list_products.
Three redundant health-check tools (check_*_service_health) perform identical operations across different services. Consider consolidating into a single 'check_service_health(service_name)' tool with an enum parameter to reduce tool proliferation and cognitive load.
No tool annotations (readOnlyHint, idempotentHint) visible in source. All tools are read-only and idempotent, but this is not explicitly declared in the MCP schema, forcing LLMs to infer safety properties.