Canvas Medical MCP by usefulapi

Each tool occupies a clearly distinct retrieval niche: capability discovery, single-resource read by id, collection search, and patient aggregate. The only potential overlap is search vs. patient-everything, but the description explicitly guides the agent to search when a resource type is known.

All names use lowercase snake_case with a consistent canvas_ prefix. The verbs vary (get, read, search) for generic vs. specific operations, which is a minor deviation but remains readable and follows FHIR conventions.

Four tools form a compact, well-scoped surface for querying FHIR records. Each tool earns its place by covering a distinct access pattern, and the count sits comfortably within the recommended 3-15 range.

For a read-oriented FHIR MCP, the surface covers capability discovery, single-resource reads, collection search, and patient $everything. Write/update operations and advanced operations like bulk export or batch bundles are absent, but core read workflows are covered.