An offline POS that saves every sale on the device first, keeps a queue of what still needs sending, and syncs it once the connection is back, without ringing anything up twice.
Each operation goes into a local queue on the device before anything else, so a dropped connection never has to stop the order in front of you.
Orders and payments keep going at zero bars
Loyalty points earned and redeemed offline
A clear status for everything still waiting to sync
Back online, exactly once
When the connection returns, waiting operations are sent in order. The server checks each one’s ID, so a sale that was already recorded is never applied a second time.
Retries that remember what already went through
Failed attempts tracked and retried
Reports that show how fresh each till’s data is
Honest about the edges
What works offline depends on your devices and setup, and we walk through yours in the demo. Card authorisation and other outside services still need their own connection.
Hospitality up front. Engineering underneath.
Designed around interrupted connections, busy teams, and the details that make every shift add up.
How offline sync works
Local storage
Save on the device
Operations enter a local queue, so a connection interruption does not have to interrupt the workflow.
Tracked operations
Reconnect and retry
Pending operations are sent when connectivity returns. Failed attempts are tracked for retry.
Duplicate checks
Check before applying
The server checks operation identifiers and returns an existing successful result when an operation is replayed.
Simplified workflow of the device application. Not a product screenshot.
Continuity, built into the deviceA local queue, not just a connection indicator.
The device application stores pending operations in SQLite with their status and retry history. Synchronization processes that queue separately. Offline capabilities depend on the device and setup; we walk through yours in the demo.
Retries with a memoryReconnecting should not mean repeating a sale.
The server checks an operation log scoped to the restaurant, device, and operation ID. Previously successful operations return their recorded result instead of applying the business action again. Failed operations can be retried; this is not a promise that every conflict resolves automatically.
Access with clear boundariesPermissions are checked in the restaurant’s context.
Server-side permission checks verify the restaurant scope, then evaluate user-specific or role-based permissions for reading, writing, and deleting resources. This is a product mechanism, not a security certification.
AI with guardrailsKept apart from your data, and from your money.
The AI runs in its own function with no access to the database. It sees only what the person asking is allowed to see, its figures come from your reports, and any change it proposes is limited to small menu and printer changes that wait for your confirmation. Each restaurant has a monthly AI allowance and spending cap.
We turn the Wi-Fi off mid-sale. On purpose.
Book a demo and we ring up a full order with the connection cut, then reconnect and watch it sync.