Sage 50 and Adobe Commerce integration

What moves between them

The main flow runs from Adobe Commerce into Sage 50. ml-connector reads completed Adobe Commerce orders and the invoices captured against them and writes them into Sage 50 as sales invoices, matched to the right customer and inventory items. New storefront shoppers are created as Sage 50 customer records so each invoice references an existing account. Adobe Commerce credit memos post into Sage 50 as the corresponding refund or credit entries. Inventory item references are aligned so invoice lines map to valid Sage 50 stock codes. Sage 50 is the accounting system of record, so ml-connector does not write financial postings back into Adobe Commerce.

How ml-connector handles it

Because Sage 50 has no cloud API, a lightweight Windows agent runs on the customer machine as a scheduled task or tray application, holding the Sage 50 SDK session and the dedicated Sage username and password, and syncing to the cloud side over HTTPS. ml-connector stores both credential sets encrypted, signs each Adobe Commerce REST call with OAuth 1.0a on PaaS or refreshes the IMS bearer token before its roughly 24-hour expiry on the Cloud Service, and verifies the HMAC-SHA256 signature on any Adobe Commerce webhook so an order-placed or invoice-saved event can trigger the sync. Adobe Commerce is polled with searchCriteria paging as the reliable backstop, since Sage 50 itself cannot be pushed to. Before creating a Sage 50 record the agent looks up the existing customer, invoice number, or stock code, because the SDK has no idempotency key, and it stores the Sage internal ID in a mapping table to prevent duplicates. The agent respects the single-session lock, so the integration user must differ from anyone logged into Sage 50 interactively, and the agent SDK version must match the installed Sage 50 year version. Failed writes are retried and every record carries a full audit trail with error replay.

A real-world example

A specialty retailer with around 60 staff sells through an Adobe Commerce B2C storefront and keeps its books in Sage 50 on a back-office Windows PC. Before the integration, a bookkeeper exported the day's web orders and re-keyed each one into Sage 50 as a sales invoice, set up new web buyers as customers by hand, and reconciled refunds separately, which left the ledger a day or two behind the store and prone to typos on totals and SKUs. With Sage 50 and Adobe Commerce connected, each completed order and its invoice flow into Sage 50 automatically against the right customer and stock items, new shoppers are created as customers on first order, and refunds post as matching entries. The books track the storefront the same day and the manual re-keying is gone.

What you can do

• •Write completed Adobe Commerce orders and invoices into Sage 50 as sales invoices against matching customers and inventory items.
• •Create new Adobe Commerce storefront shoppers as Sage 50 customer records so every invoice references a valid account.
• •Post Adobe Commerce credit memos into Sage 50 as the corresponding refund entries.
• •Bridge the Sage 50 local SDK login with Adobe Commerce OAuth 1.0a or IMS OAuth 2.0, refreshing tokens before they expire.
• •Trigger on HMAC-signed Adobe Commerce webhooks and fall back to scheduled polling, with retries and a full audit trail on every record.

Questions

Which direction does data move between Sage 50 and Adobe Commerce?

The main flow is from Adobe Commerce into Sage 50. Completed orders, captured invoices, new customers, and credit memos move from the storefront into the accounting books. Sage 50 is the system of record for accounting, so ml-connector does not write financial postings back into Adobe Commerce.

Does Sage 50 need software installed for the integration to work?

Yes. Sage 50 is desktop software with no cloud REST API, so a small Windows agent runs on the machine where Sage 50 is installed and uses the local SDK to read and write records. The agent runs as a scheduled task or tray application and syncs to the cloud side over HTTPS, which means that machine needs to stay on and logged in for syncs to run.

How are duplicate invoices avoided when Sage 50 has no idempotency key?

The Sage 50 SDK does not accept an external ID for deduplication, so before creating a record the agent first looks up an existing customer, invoice number, or stock code. It then stores the Sage-assigned internal ID in a mapping table, and on the connector side BullMQ job IDs guard against reprocessing the same Adobe Commerce event twice.