Acumatica and Deel integration

What moves between them

The main flow runs from Deel into Acumatica. ml-connector reads Deel HRIS people and writes corresponding Employee records into Acumatica, so headcount and departments reflect Deel hires, terminations, and rehires. After each billing cycle it reads Deel billing invoices and contractor payments and posts the workforce cost into Acumatica as journal transactions or AP bills, mapped to the matching GL accounts. Reference data such as departments and GL accounts is aligned so each cost line lands on a valid Acumatica account. Acumatica is the system of record for finance, so ml-connector does not write GL postings back into Deel.

How ml-connector handles it

ml-connector stores both credential sets encrypted. On the Deel side it sends the Bearer organization token and adds the x-client-id header when OAuth2 is used, and verifies the HMAC-SHA256 webhook signature against the raw request body before processing any event. On the Acumatica side it requests an OAuth 2.0 token from the tenant identity server, accepts the full instance URL per customer since Acumatica publishes no shared base URL, and pins the endpoint version to the customer's ERP release so paths resolve. Deel hires and terminations can arrive by webhook for near real-time updates, while invoice and payment data is polled on a schedule tied to your billing calendar. Departments and GL accounts are mapped first so every Employee and cost line references a value that already exists in Acumatica, and field values are wrapped in the value objects Acumatica requires. Writes to Acumatica use upsert checks on natural keys to avoid duplicates, and both Deel and Acumatica return HTTP 429 when throttled, so ml-connector backs off with jitter and retries, with a full audit trail on every record.

A real-world example

A 400-person software company runs Acumatica Cloud ERP for finance and uses Deel to employ engineers and contractors across a dozen countries through employer-of-record and contractor agreements. Before the integration, the finance team downloaded Deel billing invoices each month, sorted the labor cost by team, and keyed the totals into Acumatica journals by hand, while HR separately updated the Acumatica employee list from a spreadsheet. With Acumatica and Deel connected, new hires and leavers flow into the Acumatica Employee list, and each cycle's Deel invoices post as journal entries allocated to the right GL accounts and departments. Month-end close starts with workforce cost already booked, and the manual export and re-keying steps are gone.

What you can do

• •Keep the Acumatica Employee list aligned with Deel hires, terminations, and rehires across all countries.
• •Post workforce cost from Deel billing invoices into Acumatica as journal transactions or AP bills.
• •Map Deel departments and cost data to valid Acumatica GL accounts so every line lands on a real account.
• •Bridge Deel Bearer token auth and Acumatica OAuth 2.0, pinning the endpoint version to the customer's ERP release.
• •Take Deel employee events by webhook and poll invoices on your billing schedule, with retries and a full audit trail.

Questions

Which direction does data move between Acumatica and Deel?

The main flow is Deel into Acumatica. HRIS people, billing invoices, and contractor payments move from Deel into Acumatica, where they become Employee records and journal or AP postings. Acumatica is the finance system of record, so ml-connector does not write general ledger entries back into Deel.

Does the integration use Deel webhooks or polling?

It uses both. Deel sends signed webhooks for events such as hris.employee.created and hris.employee.terminated, which ml-connector verifies with the HMAC-SHA256 signature against the raw body for near real-time headcount updates. Invoice and payment data is polled on a schedule tied to your billing calendar, since that data settles on a cycle rather than per event.

How does ml-connector handle Acumatica's version-locked endpoints?

Acumatica endpoint paths must exactly match the running ERP version or the call returns 404. ml-connector accepts the full tenant URL and the endpoint version per customer and uses them on every request. It also wraps field values in the value objects Acumatica requires and uses upsert checks on natural keys so repeated writes do not create duplicates.