Confirm the FHIR server's Bulk Data implementation supports the _typeFilter parameter by consulting server documentation or CapabilityStatement
Construct the $export kick-off URL: GET /Patient/$export?_type=Observation&_typeFilter=Observation%3Fcode%3Dhttp%3A%2F%2Floinc.org%7C{LOINC-code} with Prefer: respond-async
URL-encode the _typeFilter value, which is a FHIR search parameter string (e.g., Observation?code=http://loinc.org|{code}); the filter is applied per resource type listed in _type
Poll the Content-Location URL until the export completes; download and parse the NDJSON output, verifying each Observation.code matches the expected LOINC code
If filtering by date is also needed, combine _typeFilter with _since (last-updated date) to further narrow the export scope
Known gotchas
_typeFilter is marked as optional in the Bulk Data IG and is not universally implemented; servers that do not support it will either ignore the filter (returning all resources of that type) or return a 400 error
The _typeFilter value must use FHIR search parameter syntax, not FHIR path or FHIRPath; using the wrong syntax silently falls back to no filtering on servers that do partial parsing
Combining _typeFilter with _type is required; specifying _typeFilter without a matching _type entry for the same resource type will result in the filter being ignored or a server error
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp