Microsoft Dynamics 365 F&O and Plaid integration

What moves between them

The flow runs from Plaid into Microsoft Dynamics 365 F&O. ml-connector reads cleared and pending bank transactions and refreshed account balances from Plaid and posts them into D365 as general journal entries mapped to the matching bank and GL accounts for reconciliation. Account and routing numbers from Plaid Auth are read to verify vendor bank details before vendor payment journal lines are released in D365. Plaid is treated as a read-only banking source for accounting, so ml-connector does not post ledger entries or vendor records back into Plaid. Transaction pulls are driven by the Plaid sync cursor, and a scheduled poll backfills anything a webhook missed.

How ml-connector handles it

ml-connector stores both credential sets encrypted. On the Plaid side it holds the client ID and secret plus the per-user access token from the Link exchange, sends every request as a POST with the Plaid-Version header, and pulls transactions through the cursor-based sync endpoint so only added, modified, and removed records move each run. On the D365 side it requests an OAuth2 client credentials token from Microsoft Entra ID scoped to the tenant environment host, refreshes it on expiry, and writes postings through journal entities because posted GL entries are read-only. Each Plaid account is mapped to a D365 bank and GL account first, so every journal line references an account that already exists. The Plaid SYNC_UPDATES_AVAILABLE webhook, verified by fetching the rotating JWK with the JWT key id, triggers a sync; ITEM ERROR and PENDING_EXPIRATION webhooks surface bank connections that need re-authentication. Plaid returns HTTP 429 on per-Item and per-client limits and a shared institution rate limit, and D365 returns 429 with a Retry-After header, so ml-connector backs off and retries on both. Because each Plaid bank connection requires the user to complete Link in a browser, there is no bulk import of bank credentials.

A real-world example

A mid-sized property management firm with roughly 150 staff runs Microsoft Dynamics 365 F&O for accounting across several operating entities and holds rent collection and operating accounts at three banks. Before the integration, an accountant downloaded statements from each bank portal every morning and keyed deposits and fees into D365 by hand, and bank reconciliation lagged a week behind. With Plaid connected, each morning's cleared transactions and refreshed balances flow into D365 as journal entries against the right bank accounts, and verified account and routing numbers confirm new vendor banking before payment runs. Reconciliation is current daily and the manual statement download is gone.

What you can do

• •Post cleared and pending Plaid bank transactions into Microsoft Dynamics 365 F&O general journal entries against the matching bank and GL accounts.
• •Pull refreshed Plaid account balances into D365 to support daily cash and bank reconciliation.
• •Verify vendor account and routing numbers from Plaid Auth before D365 vendor payment journal lines are released.
• •Bridge the Plaid client ID and per-user access token to the D365 OAuth2 token on the tenant environment host.
• •Sync only changed transactions by Plaid cursor on webhook or schedule, with retries and a full audit trail on every record.

Questions

Which direction does data move between Microsoft Dynamics 365 F&O and Plaid?

The flow runs from Plaid into Microsoft Dynamics 365 F&O. Bank transactions, balances, and account and routing numbers are read from Plaid and used to create journal entries and verify vendor banking in D365. Plaid is a read-only banking source for accounting, so ml-connector does not write ledger entries or vendor records back into Plaid.

How does ml-connector bridge the two different authentication schemes?

Plaid uses a client ID and secret plus a per-user access token from the Plaid Link flow, while Microsoft Dynamics 365 F&O uses OAuth2 client credentials through Microsoft Entra ID scoped to a tenant-specific host. ml-connector stores both credential sets encrypted, sends Plaid calls as POST with the per-user access token, and requests and refreshes the D365 token automatically as it expires.

Does the integration use Plaid webhooks or polling?

Both. ml-connector listens for the Plaid SYNC_UPDATES_AVAILABLE webhook, verified with the ES256 JWT and the rotating JWK key, and then pulls only changed transactions through the cursor-based sync endpoint. A scheduled poll backfills anything a webhook missed, and ITEM error webhooks flag bank connections that need the user to re-authenticate through Link.