FRED via OMNI Client API
OMNI exposes a small, opinionated FRED surface that is stable enough to build on:- REST endpoints under
/v1/fred/* - Equivalent tool access via hosted MCP (
fred.search,fred.series)
REST vs hosted MCP: which should you use?
Use REST if you:- want simple HTTP calls with predictable request/response envelopes
- are not using MCP tooling
- are integrating OMNI into an MCP toolchain
- want tool schemas and JSON-RPC interoperability
Required scope
All FRED access requires:Search for a series
REST: GET /v1/fred/search
Query params:
q(required): keyword querylimit(optional):1-50(default10)
missing_q(400): you omittedqinsufficient_scope(403): your key is missingfred.read
Hosted MCP: tools/call with fred.search
Fetch series observations
REST: GET /v1/fred/series/{series_id}
Path:
series_id(required): exampleCPIAUCSL
observation_start(optional):YYYY-MM-DDobservation_end(optional):YYYY-MM-DDlimit(optional): max number of observations (default1000)
Hosted MCP: tools/call with fred.series
Stability and compatibility notes
- Tool identifiers (
fred.search,fred.series) are treated as stable. - Adding optional fields is considered non-breaking.
- Any breaking change will ship as a new identifier (example:
fred.series.v2) with a sunset window for the old tool.
- the generated reference (
Client API (Generated)), and GET /v1/openapi.json