Skip to main content
Webhooks are how Maple pushes things to you — a new order, a status change, a finished menu sync. You register an HTTPS endpoint and the event types you care about; Maple signs each event and delivers it, retrying on failure and keeping a ledger you can replay from. You never poll.

The event envelope

Every delivery has the same Stripe-style envelope. The data payload is event-specific:
string
Unique event id. Dedupe on this — delivery is at-least-once.
string
The event type. Switch on it to route the event.
number
Unix seconds when the event was created.
object
Event-specific payload. order.notification carries the full order content; the lifecycle events (order.created, order.paid, order.cancelled) carry summary fields. See Webhook events for each payload, and fetch GET /v1/orders/{orderId} for authoritative current state.

Event types

Subscribe only to what you act on. The live catalog is always at GET /v1/webhook_event_types: webhook.test is also delivered on demand by the test endpoint, so you can verify a handler before any real traffic.
These are discrete events, not a status feed. Maple doesn’t send a webhook for every order status change — the transitions you drive yourself (accept, ready, complete) aren’t echoed back, and there’s no per-transition event. For an order’s current state, read GET /v1/orders/{orderId}. See the Order lifecycle.
For what each order event carries and which to subscribe to, see Webhook events.

Subscribing

Response
The signing_secret (mwhsec_…) is returned once, at creation, and never again. Store it securely now. If you lose it, rotate by creating a new subscription.
The notification_url must be public HTTPS — private and internal addresses are rejected. Manage subscriptions with GET, PATCH (update the URL, event types, or enabled/disabled status), and DELETE on /v1/webhook_subscriptions/{id}. Updating a subscription leaves its signing secret unchanged.

Multiple subscriptions

You can register more than one subscription. Every enabled subscription whose event_types include a fired event gets its own signed delivery, each with its own signing secret — so overlapping event_types across subscriptions are allowed. This lets you fan out by destination: for example, point order events at your fulfillment service and menu.sync.* events at a separate back-office endpoint. Replay follows the same routing and skips any subscription that already received the event.

Verifying the signature

Each delivery carries two headers:
  • maple-webhook-id — the event id.
  • maple-webhook-signaturet=<unix_seconds>,v1=<hex_hmac>.
The HMAC-SHA256 signature is computed over a signed string that binds the timestamp, the subscription, and the destination URL to the body — so a captured signature can’t be replayed against a different subscription or URL:
Recompute it with your signing secret and compare in constant time. Always use the raw, unparsed request body — re-serializing JSON will change the bytes and break the signature.

Delivery guarantees

  • At-least-once. The same event can arrive more than once. Dedupe on the envelope id and make your handler idempotent.
  • Respond fast. Return a 2xx quickly, then do slower work asynchronously. Any non-2xx (or a timeout) counts as a failed delivery.
  • Order is not guaranteed. Don’t assume events arrive in the order they occurred; reconcile against the order resource when sequence matters.

Retries and auto-disable

Each event is delivered with up to 9 attempts — the first immediately, then with increasing backoff. The early attempts fit inside the six-minute POS decision window, while the long tail supports reconciliation after a real outage: A non-2xx response (or a timeout — the per-attempt limit is 15 seconds) fails that attempt. An event that fails all 9 attempts counts as one fully-failed delivery. After 5 consecutive fully-failed deliveries — five separate events that each exhausted every attempt — the subscription is automatically disabled; a single successful delivery resets the counter to zero. Re-enable it with a PATCH setting status back to enabled once your endpoint is healthy, then replay anything you missed.

The event ledger

Every event is recorded, so a missed or failed delivery is recoverable.
  • GET /v1/webhook_events — your recent events, most recent first (up to 50). Useful for reconciliation and debugging.
  • GET /v1/webhook_events/{eventId} — a single event, including its full data payload.
  • POST /v1/webhook_events/{eventId}/replay — re-deliver the event to currently-matching subscriptions. Idempotent: subscriptions that already received it are skipped, so replaying is safe.
If your service was down long enough to miss an order, use the ledger replay first when you want the original event identity. For a direct catch-up, use since as your last successful checkpoint and drain forward: request GET /v1/orders?since=<checkpoint>&limit=100, then repeat with starting_after=<last-order-id> while has_more is true. With since, results are oldest first, so this reaches the entire backlog without skipping orders that share a timestamp. Fetch the authoritative order resource, then call POST /v1/orders/{orderId}/resend only for a live order that missed its webhook. Resend creates a fresh signed order.notification with a new event identity and never changes order state. Terminal orders (REJECTED, CUSTOMER_CANCELLED, or STORE_CANCELLED) cannot be resent or revived; reconcile them with GET /v1/orders?since=... and GET /v1/orders/{orderId}.

Testing your handler

This delivers a signed webhook.test event and returns whether it was delivered and the HTTP status your endpoint returned:
Use it to confirm — before any real order — that your endpoint is reachable, verifies the signature, and returns 2xx.

Checklist

Verify the HMAC signature on every delivery against the raw body, and reject stale timestamps.
Dedupe on the envelope id; make handlers idempotent.
Return 2xx fast; process asynchronously.
Subscribe only to the event types you handle.
Monitor for auto-disabled subscriptions and replay from the ledger after an outage.