Oracle Fusion Cloud ERP and SFTP / Flat Files integration

What moves between them

ml-connector reads journal batches, journal headers, and journal lines from Oracle Fusion on a polling schedule tied to your close cycle. It formats each journal entry as a CSV row with the GL account, cost center, debit or credit amount, and effective date. Invoices and purchase orders are also available as CSV. Once formatted, the connector deposits the file to your SFTP server's outbound folder. The direction is one-way out of Oracle Fusion; updates from the SFTP server back to Oracle Fusion require a separate reverse flow.

How ml-connector handles it

ml-connector stores your Oracle Fusion OAuth 2.0 credentials encrypted, refreshes the bearer token every 55 minutes before it expires, and filters API calls by LastUpdateDate using a rolling timestamp window so only new or changed records are pulled on each poll. For SFTP, it stores your SSH private key or password encrypted, initiates a new SSH connection on each poll, and uploads the formatted CSV file to the outbound folder with a timestamp to avoid collisions. The file format (CSV column order, X12 segment structure, or fixed-width layout) is configured per integration; the connector maps Oracle Fusion GL accounts to your chosen output columns. If an SFTP upload fails due to network timeout or server rejection, the connector retries with exponential backoff. Every journal entry and invoice is logged with a unique record ID, timestamp, and byte count, so you can audit which records left Oracle Fusion and verify they arrived on SFTP.

A real-world example

A mid-sized manufacturing company runs Oracle Fusion Cloud ERP for procurement and finance, and uses a legacy on-premise accounting system for consolidation and reporting that accepts data only via SFTP CSV imports. The procurement team creates purchase orders in Oracle Fusion, but the legacy system must receive PO headers and line items daily via flat file to keep buyer assignments and budgets aligned. Before the integration, the procurement coordinator exported a CSV report from Oracle Fusion each morning and uploaded it manually to the SFTP server. With Oracle Fusion and SFTP / Flat Files connected, each purchase order created or changed in Oracle Fusion automatically flows to the legacy system within minutes, and the manual export-upload step is eliminated.

What you can do

• •Extract journal entries, invoices, and purchase orders from Oracle Fusion and format them as CSV or X12 EDI files.
• •Authenticate to Oracle Fusion with OAuth 2.0 Client Credentials and handle token refresh before expiry.
• •Authenticate to your SFTP server using SSH key or password, with encrypted credential storage.
• •Poll Oracle Fusion on your schedule, filtering by LastUpdateDate so only new or changed records are synced.
• •Track which files were uploaded and log every record movement with full audit trail for reconciliation and compliance.

Questions

Does the SFTP side need to send data back into Oracle Fusion?

This integration pulls journal entries and invoices from Oracle Fusion and pushes them to SFTP as outbound files; it is one-way. If your legacy system or SFTP partner needs to send updates back into Oracle Fusion (for example, updated cost allocations or exception handling), that requires a separate reverse flow configured as its own integration.

How does ml-connector avoid sending the same record twice?

ml-connector filters the Oracle Fusion API call by LastUpdateDate, capturing only records created or modified since the last successful poll. When the SFTP upload completes, it records the file name and the record IDs it contained. On the next poll, any record already in that file is skipped, and only truly new or changed records are exported.

What happens if the SFTP server is unreachable when ml-connector tries to upload?

ml-connector retries the SFTP connection with exponential backoff over several minutes. If the upload eventually succeeds, the file is placed in the outbound folder and logged. If the SFTP server remains unavailable after the retry window, ml-connector records the failure and alerts you so the file can be uploaded manually or the outage investigated.