TallyPrime and Stampli integration

What moves between them

The main flow moves invoices and purchase orders from TallyPrime into Stampli. ml-connector polls TallyPrime's voucher and purchase order exports at a configurable interval (typically 5-15 minutes), normalizes them into CSV files, and uploads them to Stampli's SFTP inbound directory. Payment records and GL account mappings that Stampli exports are polled from the outbound directory and imported back into TallyPrime as Payment and Receipt vouchers, keeping the ledger synchronized with AP execution.

How ml-connector handles it

ml-connector runs a local agent on the same LAN as the TallyPrime instance to bridge cloud connectivity to the HTTP server on port 9000. It polls TallyPrime's Export Data endpoint using date range filters (YYYYMMDD format) to detect new invoices and purchase orders, tracks last-seen record IDs to avoid re-processing, and transforms the results into normalized CSV files. On the Stampli side, ml-connector maintains SFTP credentials encrypted and uploads files to Stampli's inbound directory, then polls the outbound directory on a schedule to detect payment and GL account exports. Because TallyPrime is single-user and does not support concurrent requests, ml-connector keeps all calls sequential. Duplicate Import Data calls to TallyPrime create duplicate vouchers, so the Stampli payment imports use Alter actions for updates where the voucher already exists. Stampli enforces duplicate detection internally, so ml-connector deduplicates on the receiving side. Every poll stores a checkpoint so a failed downstream import can be replayed without losing data.

A real-world example

A mid-sized business in South Asia runs TallyPrime for accounting and inventory across multiple locations. The finance team receives vendor invoices by email and enters them manually into TallyPrime, then exports lists to a spreadsheet for AP approval and tracking. With TallyPrime and Stampli connected, purchase orders and invoices flow automatically from TallyPrime into Stampli's approval workflows, eliminating manual data entry and enabling AI-powered invoice matching and duplicate detection. Once approved and paid, payment records flow back into TallyPrime, keeping the accounting ledger in sync with actual cash outflows without manual reconciliation.

What you can do

• •Poll TallyPrime invoices and purchase orders on a schedule, and deliver them to Stampli via SFTP in normalized CSV format.
• •Import payment and GL account records from Stampli back into TallyPrime's ledger as Payment and Receipt vouchers.
• •Store and present TallyPrime HTTP credentials and Stampli SFTP credentials encrypted, with support for optional company-level authentication on the TallyPrime side.
• •Handle TallyPrime's single-user architecture by sequencing all requests, tracking last-seen record IDs to avoid duplicates, and implementing Alter actions for updates.
• •Maintain a full audit trail of every invoice, purchase order, and payment record with checkpoints for replay in case of downstream failure.

Questions

How does ml-connector access TallyPrime if it is desktop-only and runs on a local machine?

ml-connector runs a local agent on the same LAN as the TallyPrime instance. The agent bridges cloud connectivity to TallyPrime's HTTP server on port 9000, which must be manually enabled in TallyPrime's Advanced Configuration. All requests go through the local agent, so TallyPrime never needs direct internet access.

Does Stampli support a direct REST API, or is file-based integration the only option?

File-based integration via SFTP is the only option available to third-party developers without a formal Stampli partnership agreement. Stampli does not publish a public REST API. ml-connector uploads CSV files to Stampli's SFTP inbound directory and polls the outbound directory for exports, with typical refresh rates of 5 minutes for invoices and GL accounts.

What happens if TallyPrime or Stampli rejects a record during import?

ml-connector stores checkpoints after every successful poll, so rejected records can be replayed without data loss. TallyPrime's Import Data operation returns error information in LINEERROR or IMPORTRESULT blocks, which are captured in the audit trail. Failed imports are held in a queue until they are resolved or manually reviewed.