DATEV and Procore integration

What moves between them

The flow runs from Procore into DATEV. ml-connector reads approved subcontractor invoices, direct costs, and committed purchase orders from Procore and writes them into DATEV as EXTF CSV booking batches, with each line mapped to the right DATEV debit and credit account, tax code, and cost center. The matching invoice PDFs are uploaded into DATEV Unternehmen Online under the correct client-specific document type so the source paperwork sits beside the booking. Procore vendors are carried into the DATEV creditor master rows that the booking file references. Because posted entries and the chart of accounts cannot be read from DATEV, ml-connector treats DATEV as a write-only accounting destination and never pulls bookings back into Procore.

How ml-connector handles it

ml-connector stores both credential sets encrypted and runs two separate auth paths. On the Procore side it requests a client-credentials token for the service account and refreshes it before the 90 minute expiry, paginating each project at up to 100 records per page and reading the rate-limit headers to back off on a 429. On the DATEV side it holds the PKCE-obtained refresh token from the tax advisor's one-time login and mints a fresh 15 minute access token for every batch, sending client_id only on refresh as DATEV requires. Procore cost codes and cost types are mapped first to DATEV accounts and cost centers, so every EXTF line lands on an account that already exists, since DATEV will not return its chart of accounts to validate against. EXTF files are written as UTF-8 with precomposed characters and deterministic filenames, because DATEV rejects duplicate filename and document-type pairs, which makes a re-submission safe. After upload ml-connector polls the DATEV job endpoint with backoff until the batch reports complete or failed, since there is no push notification, and a Procore Requisitions or DirectCosts webhook can trigger a run while a nightly poll reconciles anything a webhook missed. Every record carries a full audit trail and can be replayed if a downstream call fails.

A real-world example

A mid-sized German general contractor with around 200 staff runs project budgets, commitments, and subcontractor billing in Procore across several active job sites, while its tax advisor keeps the statutory books in DATEV. Before the integration, an accounts clerk exported approved subcontractor invoices and direct costs from each Procore project every week and typed the totals into DATEV by hand, then emailed the invoice PDFs separately, and month-end close stalled while the team reconciled project costs against the ledger. With DATEV and Procore connected, each approved invoice and direct cost posts into DATEV as a booking line on the right account and cost center, the PDF arrives in DATEV Unternehmen Online attached to the same client, and the manual export and re-keying step is gone.

What you can do

• •Post approved Procore subcontractor invoices and direct costs into DATEV as EXTF booking batches on the cost center for each project.
• •Upload the matching Procore invoice PDFs into DATEV Unternehmen Online under the correct client-specific document type.
• •Map Procore cost codes and cost types to DATEV accounts and tax codes so every booking line lands on a valid account.
• •Bridge Procore client-credentials tokens and the interactive DATEV PKCE login, refreshing each before expiry.
• •Trigger runs from Procore webhooks, poll DATEV job status to completion, and replay any record after a failed downstream call.

Questions

Which direction does data move between DATEV and Procore?

The flow runs from Procore into DATEV. Approved subcontractor invoices, direct costs, committed purchase orders, and their invoice PDFs move from Procore into DATEV. DATEV does not return posted journal entries or its chart of accounts through the API, so ml-connector treats DATEV as a write-only destination and never pushes bookings back into Procore.

How does the integration handle DATEV's lack of webhooks and async file jobs?

DATEV writes bookings as asynchronous EXTF or DXSO file jobs and provides no outbound webhooks. ml-connector submits each batch, receives a job ID, and polls the DATEV job status endpoint with exponential backoff until it reports complete or failed. On the Procore side, create and update webhooks can trigger a run while a nightly poll reconciles anything a webhook missed.

How are Procore cost codes mapped to DATEV accounts?

Procore organizes financial data by cost code and cost type, while DATEV bookings require explicit debit and credit account numbers and tax codes. ml-connector maps each Procore cost code and cost type to the matching DATEV account and cost center before any batch is written. This mapping is configured up front because DATEV will not return its chart of accounts for the connector to validate against.