Infermedica MCP by usefulapi

The 8 read tools split cleanly into distinct resources (condition/symptom/lab_test/risk_factor) with get vs list variants. The 9 compute endpoints (diagnosis, explain, rationale, triage, suggest, recommend_specialist, parse, search_concepts) are largely distinct but explain vs rationale and diagnosis vs triage could be confused at a glance, though descriptions disambiguate them.

Every tool carries the infermedica_ prefix and follows snake_case verb_noun conventions (get_condition, list_lab_tests, recommend_specialist, search_concepts). The compute verbs are single actions (parse, triage, suggest) but the pattern is uniform and predictable.

17 tools is on the heavier side but justified: four medical-model resources each need get+list (8 tools), plus the distinct engine operations. No tool feels redundant, though it sits near the upper bound of the comfortable 3-15 range.

The surface closely mirrors the Infermedica Engine API — full read coverage of the model plus all major compute endpoints (diagnosis, triage, explain, rationale, suggest, recommend_specialist, parse, search). Stateless compute means no update/delete is expected, so coverage is essentially complete with only marginal gaps.