Skip to main content
Writes go to the same company-scoped route as reads:
The payload goes under data:
Which data models support writes depends on the integration. Attempting an unsupported write returns 400 naming the integration and data model. You can check field-level support in our coverage tool.
Writes require a healthy connectionIf the connection is expired or disconnected the write is rejected before it reaches the platform. See Error Handling.

​
Synchronous writes

Writes are synchronous by default. Latency depends on the integration — we add only the overhead of mapping and logging.
Some calls take much longer than othersThis comes from the integration itself: a single unified write sometimes requires several calls to the platform.

​
Asynchronous writes

Pass run_async=true as a query parameter:
The response returns a sync_id. Track it with GET /v4/core/companies/{companyId}/syncs, or subscribe to the SYNC_COMPLETED webhook. We recommend async for large datasets — rate limiting and retries are handled for you.

​
Batch writes

Send data as an array of objects. All records in the batch share one sync_id.
Batches require run_async=trueSending more than one record with run_async=false (or omitted) returns 400. Either pass run_async=true or send one record per request.
One data model per request. To write 10 invoices and 10 bills, call the invoices endpoint and the bills endpoint separately.

​
Uploading documents

Documents use a dedicated endpoint, POST /v4/core/companies/{companyId}/documents, and accept either a multipart file upload or a base64-encoded file in the JSON body. Each document attaches to a parent record via document_type and document_type_id.