TallyPrime and Coupa integration

What moves between them

Purchase order and invoice data flows from Coupa into TallyPrime. When Coupa fires a purchase-order-event or invoice-event, ml-connector translates the event into a TallyPrime Purchase or Purchase Order voucher and posts it to the local TallyPrime instance via the HTTP API. Supplier master data flows both directions: Coupa supplier records are polled periodically and written to TallyPrime as Ledger entries, and TallyPrime ledgers can be exported and matched against Coupa's supplier list for reconciliation. The local agent maintains deduplication state so that retried or replayed events do not create duplicate vouchers in TallyPrime.

How ml-connector handles it

ml-connector stores both Coupa OAuth credentials and TallyPrime company credentials encrypted in the database, and routes TallyPrime API calls through a local agent running on the same machine or LAN as the desktop application. When Coupa fires a webhook event, ml-connector translates it to TallyPrime's XML voucher format, deduplicates on the local side using the Coupa event ID as the idempotency key, and POSTs the request to the local port 9000 server. If TallyPrime returns an IMPORTRESULT error, ml-connector logs it and can retry the same event ID without creating a duplicate. For supplier master data, ml-connector polls Coupa's suppliers via REST and compares against last-seen state, translating new or changed suppliers into TallyPrime Ledger Import requests with the company name and optional authentication supplied by the local agent. Rate limits on the Coupa side trigger exponential backoff retries capped at 30 seconds. The local agent must be kept running on the same LAN and the TallyPrime company must be open in the application for API calls to succeed.

A real-world example

A mid-sized manufacturing distributor in Southeast Asia uses TallyPrime for accounting and inventory on Windows desktops at each branch office, and Coupa in the cloud for centralized supplier management and purchase order approval across all branches. Before the integration, purchase orders were approved in Coupa at head office, then a team member at each branch manually entered the order data into TallyPrime or received email PDFs to manually re-key. With the integration, each approved Coupa purchase order is automatically translated into a TallyPrime Purchase Order voucher in the local instance, allocated to the right expense ledger and supplier account. Supplier master changes at head office in Coupa flow down to all branches' TallyPrime instances so ledger entries always reference the current supplier. Manual re-entry is eliminated and supplier records stay aligned across the system.

What you can do

• •Automatically create TallyPrime purchase order vouchers from Coupa purchase order events, allocated to the correct expense ledger.
• •Keep TallyPrime supplier ledgers in sync with Coupa's supplier master, preventing mismatched vendor names and accounts.
• •Deduplicate purchase orders on the local agent side so retried or replayed Coupa events do not create duplicate vouchers in TallyPrime.
• •Authenticate Coupa with OAuth2 and TallyPrime with optional company-level credentials, with credentials encrypted and stored securely.
• •Maintain a persistent local HTTP tunnel to TallyPrime so cloud procurement events can reach the desktop application without manual intervention.

Questions

Does ml-connector require a local agent on the same machine as TallyPrime?

Yes. TallyPrime runs on port 9000 as a local HTTP server that is only accessible on the LAN. ml-connector includes a local agent that must run on the same machine or LAN segment as the TallyPrime instance. The agent maintains the tunnel and handles deduplication so the same Coupa event is not imported twice if a network retry occurs.

What happens if the TallyPrime application is closed or the company is not open?

API calls will fail. TallyPrime is a single-user desktop application and the target company must be open in the application for Import or Export requests to succeed. If the application closes, ml-connector will retry on a schedule, and once the application and company are open again, the pending events will be processed.

How does ml-connector handle Coupa's webhook events and TallyPrime's lack of idempotency?

ml-connector deduplicates on the local agent side using the Coupa event ID as an idempotency key. When a purchase-order-event or invoice-event arrives from Coupa's webhook, the agent checks if it has already processed that event ID. If it has, the event is skipped. If not, it translates the event to TallyPrime's XML voucher format and POSTs it to port 9000. This prevents duplicate vouchers even if Coupa retries or replays the event.