When a 202 Accepted response is returned (commonly for pay statement endpoints), treat it as a signal that the data is still being fetched from the upstream provider.
Implement a polling loop with exponential backoff: wait, then re-issue the same request until you receive a 200 response with the data.
For pay statements specifically, ensure that the payment_id referenced exists in the /payment endpoint before requesting its pay statements — a missing payment_id will also yield a 202.
Use the Finch Data Refresh endpoint to enqueue an on-demand sync if you need fresher data than the scheduled sync cadence provides.
Monitor sync status and log 202 occurrences to surface data freshness issues to downstream consumers.
Known gotchas
A 202 response body typically contains a message indicating data is being fetched — do not treat it as an empty success; parse and log the message to distinguish it from actual data.
Pay statements are larger and slower to sync than directory or employment data — design your pipeline to handle pay statement latency separately from other data types.
Finch syncs data on a provider-specific cadence; very recent payroll changes at the provider may not yet be reflected even after a successful 200 response — surface data freshness timestamps to users where possible.
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