skills/vtexdocs/ai-skills/marketplace-fulfillment

marketplace-fulfillment

Installation
SKILL.md

Fulfillment, simulation, orders & OMS follow-up

When this skill applies

Use this skill when building an External Seller integration: VTEX forwards availability, shipping, checkout simulation, and order placement to your fulfillment base URL, and you call the marketplace OMS APIs for invoice and tracking after dispatch.

  • Implementing POST /pvt/orderForms/simulation (indexation and/or checkout — with or without customer context)
  • Implementing POST /pvt/orders (order placement — create reservation; return orderId = reservation id in your system)
  • Handling POST /pvt/orders/{sellerOrderId}/fulfill (order dispatch after approval — path uses the same id you returned as orderId at placement)
  • Sending invoice notifications via POST /api/oms/pvt/orders/{marketplaceOrderId}/invoice (marketplace order id in the path)
  • Updating tracking via PATCH /api/oms/pvt/orders/{marketplaceOrderId}/invoice/{invoiceNumber}
  • Implementing partial invoicing for split shipments

Do not use this skill for:

  • Catalog or SKU synchronization (see marketplace-catalog-sync)
  • Order event consumption via Feed/Hook (see marketplace-order-hook)
  • General API rate limiting (see marketplace-rate-limiting)

Decision rules

External Seller protocol (implemented on the seller host)

  • Fulfillment simulationPOST /pvt/orderForms/simulation. VTEX calls it during product indexation and during checkout. Requests may include only items (and optionally query params), or the full checkout context: items, postalCode, country, clientProfileData, shippingData, selectedSla, etc. Without postal code / profile (typical indexation), the response must still state whether each item is available. With full context, return items, logisticsInfo (one entry per requested item), postalCode, country, and set allowMultipleDeliveries to true as required by the contract. items[].id is the seller SKU id; items[].seller is the seller id on the marketplace account.
  • Response shape — Each logisticsInfo[] row aligns with a requested item (itemIndex). Include slas[] with all delivery options (home delivery and pickup-in-point when applicable), deliveryChannels[] with per-channel stock, shipsTo, and stockBalance. SLA fields include price (shipping per item, in cents), shippingEstimate / shippingEstimateDate (e.g. 5bd, 30m). Pickup SLAs must include pickupStoreInfo (address, friendlyName, etc.).
  • SLA — The simulation handler must respond within 2.5 seconds or the offer is treated as unavailable.
  • Order placementPOST /pvt/orders with a JSON array of orders. For each order, create a reservation in your system. The response must be the same structure with an added orderId on each order: that value is your reservation id and becomes the sellerOrderId in later protocol calls (e.g. authorize fulfillment path parameter).
  • Order dispatch (authorize fulfillment) — After marketplace approval, VTEX calls POST /pvt/orders/{sellerOrderId}/fulfill where sellerOrderId equals the orderId you returned at placement. Body includes marketplaceOrderId and marketplaceOrderGroup. Convert the reservation to a firm order in your system; response body includes date, marketplaceOrderId, orderId (seller reference), receipt.

OMS APIs (seller → marketplace)

  • Send invoices via POST /api/oms/pvt/orders/{marketplaceOrderId}/invoice. Required fields: type, invoiceNumber, invoiceValue (in cents), issuanceDate, and items array. Path marketplaceOrderId is the VTEX marketplace order id, not your reservation id.
  • Use type: "Output" for sales invoices (shipment) and type: "Input" for return invoices.
  • Send tracking separately after the carrier provides it, using PATCH /api/oms/pvt/orders/{marketplaceOrderId}/invoice/{invoiceNumber}.
  • For split shipments, send one invoice per package with only the items in that package. Each invoiceValue must reflect only its items.
  • Once an order is invoiced, it cannot be canceled without first sending a return invoice (type: "Input").
  • The fulfillment simulation endpoint must respond within 2.5 seconds or the product is considered unavailable.

Architecture / data flow (high level):

VTEX Checkout / indexation          External Seller                    VTEX OMS (marketplace)
       │                                   │                                   │
       │── POST /pvt/orderForms/simulation ▶│  Price, stock, SLAs               │
       │◀── 200 + items + logisticsInfo ───│                                   │
       │                                   │                                   │
       │── POST /pvt/orders (array) ───────▶│  Create reservation               │
       │◀── same + orderId (reservation) ─│                                   │
       │                                   │                                   │
       │── POST /pvt/orders/{id}/fulfill ──▶│  Commit / pick pack               │
       │◀── date, marketplaceOrderId, ... ─│                                   │
       │                                   │── POST .../invoice ────────────────▶│
       │                                   │── PATCH .../invoice/{n} ─────────▶│

Hard constraints

Constraint: Marketplace order ID in OMS paths

Any {orderId} in /api/oms/pvt/orders/{orderId}/... MUST be the VTEX marketplace order id (OMS), not the orderId you returned at POST /pvt/orders (reservation id). Map marketplaceOrderId from the protocol (placement payload, fulfill body, or events) before calling invoice or tracking APIs.

Why this matters

Using the reservation id in OMS URLs fails to match the marketplace order; invoices and tracking never attach to the customer order.

Detection

If the same variable is used for both POST /pvt/orders response orderId and POST .../oms/.../invoice without mapping → STOP.

Correct

// reservationId from your POST /pvt/orders response; marketplaceOrderId from VTEX payload
await omsClient.post(`/api/oms/pvt/orders/${marketplaceOrderId}/invoice`, payload);

Wrong

await omsClient.post(`/api/oms/pvt/orders/${reservationId}/invoice`, payload);

Constraint: Fulfillment simulation contract and latency

The seller MUST implement POST /pvt/orderForms/simulation to return a valid items array for every request. When the request includes checkout context (e.g. postalCode, clientProfileData, shippingData), the response MUST include aligned logisticsInfo, slas for all relevant modes (including pickup when offered), and allowMultipleDeliveries: true where required. The handler MUST complete within 2.5 seconds.

Why this matters

Incomplete logisticsInfo or missing SLAs break checkout shipping selection. Slow responses mark offers unavailable and hurt conversion.

Detection

If simulation returns only items with prices but omits logisticsInfo when the request had shippingData → warn. If p95 latency approaches 2s without caching → warn.

Correct

// Pseudocode: branch on whether checkout context is present
if (hasCheckoutContext(req.body)) {
  return res.json({
    country: req.body.country,
    items: pricedItems,
    logisticsInfo: buildLogisticsPerItem(pricedItems, req.body),
    postalCode: req.body.postalCode,
    allowMultipleDeliveries: true,
  });
}
return res.json({ items: availabilityOnlyItems /* + minimal logistics if required */ });

Wrong

// WRONG: Full checkout body but response omits logisticsInfo / SLAs
res.json({ items: pricedItemsOnly });

Constraint: Order placement must return seller orderId (reservation)

POST /pvt/orders accepts a JSON array of orders. The response MUST be the same orders with orderId set on each element to your reservation identifier (seller system). VTEX uses that value as sellerOrderId in POST /pvt/orders/{sellerOrderId}/fulfill.

Why this matters

Omitting or reusing a fake orderId breaks the link between marketplace order flow and your reservation and prevents dispatch from routing correctly.

Detection

If the handler returns 200 without adding orderId, or returns a single object instead of an array → warn.

Correct

app.post("/pvt/orders", (req, res) => {
  const orders = req.body as Array<Record<string, unknown>>;
  const out = orders.map((order) => {
    const reservationId = createReservation(order);
    return { ...order, orderId: reservationId, followUpEmail: "" };
  });
  res.json(out);
});

Wrong

app.post("/pvt/orders", (req, res) => {
  createReservation(req.body);
  res.status(200).send(); // WRONG: missing orderId echo
});

Constraint: Send Correct Invoice Format with All Required Fields

The invoice notification MUST include type, invoiceNumber, invoiceValue, issuanceDate, and items array. The invoiceValue MUST be in cents. The items array MUST match the items in the order.

Why this matters

Missing required fields cause the API to reject the invoice with 400 Bad Request, leaving the order stuck in "handling" status. Incorrect invoiceValue (e.g., using dollars instead of cents) causes financial discrepancies in marketplace reconciliation.

Detection

If you see an invoice notification payload missing invoiceNumber, invoiceValue, issuanceDate, or items → warn about missing required fields. If invoiceValue appears to be in dollars (e.g., 99.90 instead of 9990) → warn about cents conversion.

Correct

import axios, { AxiosInstance } from "axios";

interface InvoiceItem {
  id: string;
  quantity: number;
  price: number; // in cents
}

interface InvoicePayload {
  type: "Output" | "Input";
  invoiceNumber: string;
  invoiceValue: number; // total in cents
  issuanceDate: string; // ISO 8601
  invoiceUrl?: string;
  invoiceKey?: string;
  courier?: string;
  trackingNumber?: string;
  trackingUrl?: string;
  items: InvoiceItem[];
}

async function sendInvoiceNotification(
  client: AxiosInstance,
  marketplaceOrderId: string,
  invoice: InvoicePayload
): Promise<void> {
  // Validate required fields before sending
  if (!invoice.invoiceNumber) {
    throw new Error("invoiceNumber is required");
  }
  if (!invoice.invoiceValue || invoice.invoiceValue <= 0) {
    throw new Error("invoiceValue must be a positive number in cents");
  }
  if (!invoice.issuanceDate) {
    throw new Error("issuanceDate is required");
  }
  if (!invoice.items || invoice.items.length === 0) {
    throw new Error("items array is required and must not be empty");
  }

  // Warn if invoiceValue looks like it's in dollars instead of cents
  if (invoice.invoiceValue < 100 && invoice.items.length > 0) {
    console.warn(
      `Warning: invoiceValue ${invoice.invoiceValue} seems very low. ` +
        `Ensure it's in cents (e.g., 9990 for $99.90).`
    );
  }

  await client.post(`/api/oms/pvt/orders/${marketplaceOrderId}/invoice`, invoice);
}

// Example usage:
async function invoiceOrder(
  client: AxiosInstance,
  marketplaceOrderId: string
): Promise<void> {
  await sendInvoiceNotification(client, marketplaceOrderId, {
    type: "Output",
    invoiceNumber: "NFE-2026-001234",
    invoiceValue: 15990, // $159.90 in cents
    issuanceDate: new Date().toISOString(),
    invoiceUrl: "https://invoices.example.com/NFE-2026-001234.pdf",
    invoiceKey: "35260614388220000199550010000012341000012348",
    items: [
      { id: "123", quantity: 1, price: 9990 },
      { id: "456", quantity: 2, price: 3000 },
    ],
  });
}

Wrong

// WRONG: Missing required fields, value in dollars instead of cents
async function sendBrokenInvoice(
  client: AxiosInstance,
  marketplaceOrderId: string
): Promise<void> {
  await client.post(`/api/oms/pvt/orders/${marketplaceOrderId}/invoice`, {
    // Missing 'type' field — API may reject or default incorrectly
    invoiceNumber: "001234",
    invoiceValue: 159.9, // WRONG: dollars, not cents — causes financial mismatch
    // Missing 'issuanceDate' — API will reject with 400
    // Missing 'items' — API cannot match invoice to order items
  });
}

Constraint: Update Tracking Promptly After Shipping

Tracking information MUST be sent as soon as the carrier provides it. Use PATCH /api/oms/pvt/orders/{marketplaceOrderId}/invoice/{invoiceNumber} to add tracking to an existing invoice.

Why this matters

Late tracking updates prevent customers from seeing shipment status in the marketplace. The order remains in "invoiced" state instead of progressing to "delivering" and then "delivered". This generates customer support tickets and damages seller reputation.

Detection

If you see tracking information being batched for daily updates instead of sent in real-time → warn about prompt tracking updates. If tracking is included in the initial invoice call but the carrier hasn't provided it yet (hardcoded/empty values) → warn.

Correct

interface TrackingUpdate {
  courier: string;
  trackingNumber: string;
  trackingUrl?: string;
  isDelivered?: boolean;
}

async function updateOrderTracking(
  client: AxiosInstance,
  marketplaceOrderId: string,
  invoiceNumber: string,
  tracking: TrackingUpdate
): Promise<void> {
  await client.patch(
    `/api/oms/pvt/orders/${marketplaceOrderId}/invoice/${invoiceNumber}`,
    tracking
  );
}

// Send tracking as soon as carrier provides it
async function onCarrierPickup(
  client: AxiosInstance,
  marketplaceOrderId: string,
  invoiceNumber: string,
  carrierData: { name: string; trackingId: string; trackingUrl: string }
): Promise<void> {
  await updateOrderTracking(client, marketplaceOrderId, invoiceNumber, {
    courier: carrierData.name,
    trackingNumber: carrierData.trackingId,
    trackingUrl: carrierData.trackingUrl,
  });
  console.log(
    `Tracking updated for marketplace order ${marketplaceOrderId}: ${carrierData.trackingId}`
  );
}

// Update delivery status when confirmed
async function onDeliveryConfirmed(
  client: AxiosInstance,
  marketplaceOrderId: string,
  invoiceNumber: string
): Promise<void> {
  await updateOrderTracking(client, marketplaceOrderId, invoiceNumber, {
    courier: "",
    trackingNumber: "",
    isDelivered: true,
  });
  console.log(`Marketplace order ${marketplaceOrderId} marked as delivered`);
}

Wrong

// WRONG: Sending empty/fake tracking data with the invoice
async function invoiceWithFakeTracking(
  client: AxiosInstance,
  marketplaceOrderId: string
): Promise<void> {
  await client.post(`/api/oms/pvt/orders/${marketplaceOrderId}/invoice`, {
    type: "Output",
    invoiceNumber: "NFE-001",
    invoiceValue: 9990,
    issuanceDate: new Date().toISOString(),
    items: [{ id: "123", quantity: 1, price: 9990 }],
    // WRONG: Hardcoded tracking — carrier hasn't picked up yet
    courier: "TBD",
    trackingNumber: "PENDING",
    trackingUrl: "",
  });
  // Customer sees "PENDING" as tracking number — useless information
}

Constraint: Handle Partial Invoicing for Split Shipments

For orders shipped in multiple packages, each shipment MUST have its own invoice with only the items included in that package. The invoiceValue MUST reflect only the items in that particular shipment.

Why this matters

Sending a single invoice for the full order value when only partial items are shipped causes financial discrepancies. The marketplace cannot reconcile payments correctly, and the order status may not progress properly.

Detection

If you see a single invoice being sent with the full order value for partial shipments → warn about partial invoicing. If the items array doesn't match the actual items being shipped → warn.

Correct

interface OrderItem {
  id: string;
  name: string;
  quantity: number;
  price: number; // per unit in cents
}

interface Shipment {
  items: OrderItem[];
  invoiceNumber: string;
}

async function sendPartialInvoices(
  client: AxiosInstance,
  marketplaceOrderId: string,
  shipments: Shipment[]
): Promise<void> {
  for (const shipment of shipments) {
    // Calculate value for only the items in this shipment
    const shipmentValue = shipment.items.reduce(
      (total, item) => total + item.price * item.quantity,
      0
    );

    await sendInvoiceNotification(client, marketplaceOrderId, {
      type: "Output",
      invoiceNumber: shipment.invoiceNumber,
      invoiceValue: shipmentValue,
      issuanceDate: new Date().toISOString(),
      items: shipment.items.map((item) => ({
        id: item.id,
        quantity: item.quantity,
        price: item.price,
      })),
    });

    console.log(
      `Partial invoice ${shipment.invoiceNumber} sent for marketplace order ${marketplaceOrderId}: ` +
        `${shipment.items.length} items, value=${shipmentValue}`
    );
  }
}

// Example: Order with 3 items shipped in 2 packages
await sendPartialInvoices(client, "vtex-marketplace-order-id-12345", [
  {
    invoiceNumber: "NFE-001-A",
    items: [
      { id: "sku-1", name: "Laptop", quantity: 1, price: 250000 },
    ],
  },
  {
    invoiceNumber: "NFE-001-B",
    items: [
      { id: "sku-2", name: "Mouse", quantity: 1, price: 5000 },
      { id: "sku-3", name: "Keyboard", quantity: 1, price: 12000 },
    ],
  },
]);

Wrong

// WRONG: Sending full order value for partial shipment
async function wrongPartialInvoice(
  client: AxiosInstance,
  marketplaceOrderId: string,
  totalOrderValue: number,
  shippedItems: OrderItem[]
): Promise<void> {
  await client.post(`/api/oms/pvt/orders/${marketplaceOrderId}/invoice`, {
    type: "Output",
    invoiceNumber: "NFE-001-A",
    invoiceValue: totalOrderValue, // WRONG: Full order value, not partial
    issuanceDate: new Date().toISOString(),
    items: shippedItems.map((item) => ({
      id: item.id,
      quantity: item.quantity,
      price: item.price,
    })),
    // invoiceValue doesn't match sum of items — financial mismatch
  });
}

Preferred pattern

Implement fulfillment simulation

Register POST /pvt/orderForms/simulation. Parse items[] (id = seller SKU, seller = seller id on marketplace). If the body has no postalCode / clientProfileData, treat as indexation: return availability (and minimal logistics if your contract requires it). If the body includes checkout fields, build logisticsInfo per itemIndex, populate slas (delivery + pickup-in-point with pickupStoreInfo when applicable), deliveryChannels, stockBalance, set country / postalCode, and allowMultipleDeliveries: true. Keep CPU and I/O bounded so you stay under 2.5s.

import { RequestHandler } from "express";

const fulfillmentSimulation: RequestHandler = async (req, res) => {
  const body = req.body as Record<string, unknown>;
  const items = body.items as Array<{ id: string; quantity: number; seller: string }>;
  const hasCheckoutContext = Boolean(body.postalCode && body.shippingData);

  const pricedItems = await priceAndStockForItems(items);

  if (!hasCheckoutContext) {
    res.json({ items: pricedItems, logisticsInfo: minimalLogistics(pricedItems) });
    return;
  }

  res.json({
    country: body.country,
    postalCode: body.postalCode,
    items: pricedItems,
    logisticsInfo: buildFullLogistics(pricedItems, body),
    allowMultipleDeliveries: true,
  });
};

function minimalLogistics(
  pricedItems: Array<{ id: string; requestIndex: number }>
): unknown[] {
  return pricedItems.map((_, i) => ({
    itemIndex: i,
    quantity: 1,
    shipsTo: ["USA"],
    slas: [],
    stockBalance: "",
    deliveryChannels: [{ id: "delivery", stockBalance: "" }],
  }));
}

function buildFullLogistics(
  pricedItems: Array<{ id: string; requestIndex: number }>,
  _checkoutBody: unknown
): unknown[] {
  // Replace with SLA / carrier rules derived from checkoutBody
  return pricedItems.map((_, i) => ({
    itemIndex: i,
    quantity: 1,
    shipsTo: ["USA"],
    slas: [],
    stockBalance: 0,
    deliveryChannels: [],
  }));
}

async function priceAndStockForItems(
  items: Array<{ id: string; quantity: number; seller: string }>
): Promise<Array<Record<string, unknown>>> {
  return items.map((item, requestIndex) => ({
    id: item.id,
    quantity: item.quantity,
    seller: item.seller,
    measurementUnit: "un",
    merchantName: null,
    price: 0,
    priceTags: [],
    priceValidUntil: null,
    requestIndex,
    unitMultiplier: 1,
    attachmentOfferings: [],
  }));
}

Implement order placement (reservation)

Register POST /pvt/orders. Body is an array. Persist each order as a reservation; respond with the same objects plus orderId (reservation key) and typically followUpEmail. That orderId is what VTEX passes as sellerOrderId on POST /pvt/orders/{sellerOrderId}/fulfill.

import { RequestHandler } from "express";

function createReservation(_order: Record<string, unknown>): string {
  return `res-${Date.now()}-${Math.random().toString(36).slice(2, 9)}`;
}

const orderPlacement: RequestHandler = (req, res) => {
  const orders = req.body as Array<Record<string, unknown>>;
  const out = orders.map((order) => ({
    ...order,
    orderId: createReservation(order),
    followUpEmail: "",
  }));
  res.json(out);
};

Implement the Authorize Fulfillment Endpoint

The marketplace calls this endpoint when payment is approved.

import express, { RequestHandler } from "express";

interface FulfillOrderRequest {
  marketplaceOrderId: string;
  marketplaceOrderGroup: string;
}

interface OrderMapping {
  /** Same id returned as orderId from POST /pvt/orders (reservation) */
  sellerOrderId: string;
  marketplaceOrderId: string;
  items: OrderItem[];
  status: string;
}

// Store for order mappings — use a real database in production
const orderStore = new Map<string, OrderMapping>();

const authorizeFulfillmentHandler: RequestHandler = async (req, res) => {
  const sellerOrderId = req.params.sellerOrderId;
  const { marketplaceOrderId, marketplaceOrderGroup }: FulfillOrderRequest = req.body;

  console.log(
    `Fulfillment authorized: reservation=${sellerOrderId}, marketplaceOrder=${marketplaceOrderId}, group=${marketplaceOrderGroup}`
  );

  // Store the marketplace order ID mapping
  const order = orderStore.get(sellerOrderId);
  if (!order) {
    res.status(404).json({ error: "Order not found" });
    return;
  }

  order.marketplaceOrderId = marketplaceOrderId;
  order.status = "fulfillment_authorized";
  orderStore.set(sellerOrderId, order);

  // Trigger fulfillment process asynchronously
  enqueueFulfillment(sellerOrderId).catch(console.error);

  res.status(200).json({
    date: new Date().toISOString(),
    marketplaceOrderId,
    // Echo seller reservation id (same as path param / placement orderId)
    orderId: sellerOrderId,
    receipt: null,
  });
};

async function enqueueFulfillment(sellerOrderId: string): Promise<void> {
  console.log(`Enqueued fulfillment for ${sellerOrderId}`);
}

const app = express();
app.use(express.json());
app.post("/pvt/orders/:sellerOrderId/fulfill", authorizeFulfillmentHandler);

Send Invoice After Fulfillment

Once the order is packed and the invoice is generated, send the invoice notification.

async function fulfillAndInvoice(
  client: AxiosInstance,
  order: OrderMapping
): Promise<void> {
  // Generate invoice from your invoicing system
  const invoice = await generateInvoice(order);

  // Send invoice notification to VTEX marketplace
  await sendInvoiceNotification(client, order.marketplaceOrderId, {
    type: "Output",
    invoiceNumber: invoice.number,
    invoiceValue: invoice.totalCents,
    issuanceDate: invoice.issuedAt.toISOString(),
    invoiceUrl: invoice.pdfUrl,
    invoiceKey: invoice.accessKey,
    items: order.items.map((item) => ({
      id: item.id,
      quantity: item.quantity,
      price: item.price,
    })),
  });

  console.log(
    `Invoice ${invoice.number} sent for order ${order.marketplaceOrderId}`
  );
}

async function generateInvoice(order: OrderMapping): Promise<{
  number: string;
  totalCents: number;
  issuedAt: Date;
  pdfUrl: string;
  accessKey: string;
}> {
  const totalCents = order.items.reduce(
    (sum, item) => sum + item.price * item.quantity,
    0
  );
  return {
    number: `NFE-${Date.now()}`,
    totalCents,
    issuedAt: new Date(),
    pdfUrl: `https://invoices.example.com/NFE-${Date.now()}.pdf`,
    accessKey: "35260614388220000199550010000012341000012348",
  };
}

Send Tracking When Carrier Picks Up

async function handleCarrierPickup(
  client: AxiosInstance,
  marketplaceOrderId: string,
  invoiceNumber: string,
  carrier: { name: string; trackingId: string; trackingUrl: string }
): Promise<void> {
  await updateOrderTracking(client, marketplaceOrderId, invoiceNumber, {
    courier: carrier.name,
    trackingNumber: carrier.trackingId,
    trackingUrl: carrier.trackingUrl,
  });

  console.log(
    `Tracking ${carrier.trackingId} sent for marketplace order ${marketplaceOrderId}`
  );
}

Confirm Delivery

async function handleDeliveryConfirmation(
  client: AxiosInstance,
  marketplaceOrderId: string,
  invoiceNumber: string
): Promise<void> {
  await client.patch(
    `/api/oms/pvt/orders/${marketplaceOrderId}/invoice/${invoiceNumber}`,
    {
      isDelivered: true,
      courier: "",
      trackingNumber: "",
    }
  );

  console.log(`Marketplace order ${marketplaceOrderId} marked as delivered`);
}

Complete Example

import axios, { AxiosInstance } from "axios";

function createMarketplaceClient(
  accountName: string,
  appKey: string,
  appToken: string
): AxiosInstance {
  return axios.create({
    baseURL: `https://${accountName}.vtexcommercestable.com.br`,
    headers: {
      "Content-Type": "application/json",
      "X-VTEX-API-AppKey": appKey,
      "X-VTEX-API-AppToken": appToken,
    },
    timeout: 10000,
  });
}

async function completeFulfillmentFlow(
  client: AxiosInstance,
  order: OrderMapping
): Promise<void> {
  // 1. Fulfill and invoice
  await fulfillAndInvoice(client, order);

  // 2. When carrier picks up, send tracking
  const carrierData = await waitForCarrierPickup(order.sellerOrderId);
  const invoice = await getLatestInvoice(order.sellerOrderId);

  await handleCarrierPickup(
    client,
    order.marketplaceOrderId,
    invoice.number,
    carrierData
  );

  // 3. When delivered, confirm
  await waitForDeliveryConfirmation(order.sellerOrderId);
  await handleDeliveryConfirmation(
    client,
    order.marketplaceOrderId,
    invoice.number
  );
}

async function waitForCarrierPickup(
  sellerOrderId: string
): Promise<{ name: string; trackingId: string; trackingUrl: string }> {
  // Replace with actual carrier integration
  return {
    name: "Correios",
    trackingId: "BR123456789",
    trackingUrl: "https://tracking.example.com/BR123456789",
  };
}

async function getLatestInvoice(
  sellerOrderId: string
): Promise<{ number: string }> {
  // Replace with actual invoice lookup
  return { number: `NFE-${sellerOrderId}` };
}

async function waitForDeliveryConfirmation(
  sellerOrderId: string
): Promise<void> {
  // Replace with actual delivery confirmation logic
  console.log(`Waiting for delivery confirmation: ${sellerOrderId}`);
}

Common failure modes

  • Treating indexation simulation like checkout. The same endpoint receives minimal bodies (no customer location) during indexation and rich bodies during checkout. Returning checkout-grade logisticsInfo for minimal calls can be unnecessary work; returning only prices for checkout calls without SLAs and logisticsInfo breaks shipping selection.

  • Omitting orderId on order placement. VTEX expects an array response echoing each order with orderId set to your reservation. Empty 200 responses or missing orderId strand the order pipeline.

  • Using reservation orderId in OMS invoice URLs. After placement, you must use marketplaceOrderId from the protocol when calling /api/oms/pvt/orders/.... Confusing the two ids produces 404 or silent failure on invoice/tracking.

  • Sending invoice before fulfillment authorization. The seller sends an invoice notification immediately when the order is placed, before receiving the Authorize Fulfillment callback from the marketplace. Payment may still be pending or under review. Invoicing before authorization can result in the invoice being rejected or the order being in an inconsistent state. Only send the invoice after receiving POST /pvt/orders/{sellerOrderId}/fulfill.

  • Not handling return invoices for cancellation. A seller tries to cancel an invoiced order by calling the Cancel Order endpoint directly without first sending a return invoice. Once an order is in "invoiced" status, it cannot be canceled without a return invoice (type: "Input"). The Cancel Order API will reject the request.

// Correct: Send return invoice before canceling an invoiced order
async function cancelInvoicedOrder(
  client: AxiosInstance,
  marketplaceOrderId: string,
  originalItems: InvoiceItem[],
  originalInvoiceValue: number
): Promise<void> {
  // Step 1: Send return invoice (type: "Input")
  await sendInvoiceNotification(client, marketplaceOrderId, {
    type: "Input", // Return invoice
    invoiceNumber: `RET-${Date.now()}`,
    invoiceValue: originalInvoiceValue,
    issuanceDate: new Date().toISOString(),
    items: originalItems,
  });

  // Step 2: Now cancel the order
  await client.post(
    `/api/marketplace/pvt/orders/${marketplaceOrderId}/cancel`,
    { reason: "Customer requested return" }
  );
}
  • Fulfillment simulation exceeding the 2.5-second timeout. The seller's fulfillment simulation endpoint performs complex database queries or external API calls that exceed the response time limit. VTEX marketplaces wait a maximum of 2.5 seconds for a fulfillment simulation response. After that, the product is considered unavailable/inactive and won't appear in the storefront or checkout. Pre-cache price and inventory data.

Review checklist

  • Does POST /pvt/orderForms/simulation handle both minimal (indexation) and checkout-shaped requests, returning logisticsInfo and slas when context is present?
  • Does POST /pvt/orders accept an array and return each order with orderId (reservation id)?
  • Does POST /pvt/orders/{sellerOrderId}/fulfill read marketplaceOrderId and marketplaceOrderGroup and match sellerOrderId to the reservation from placement?
  • Are OMS /invoice and PATCH .../invoice calls using marketplaceOrderId, not the reservation id?
  • Does the seller only begin physical fulfillment after receiving the Authorize Fulfillment callback?
  • For IO/BFF connectors: are caching and route choices aligned with vtex-io skills (simulation SLA, data scope)?
  • Does the invoice payload include all required fields (type, invoiceNumber, invoiceValue, issuanceDate, items)?
  • Is invoiceValue in cents (not dollars)?
  • Is tracking sent separately after the carrier provides real data (not hardcoded placeholders)?
  • For split shipments, does each invoice cover only its package's items and value?
  • Is cancellation of invoiced orders handled via return invoice (type: "Input") first?
  • Does the fulfillment simulation endpoint respond within 2.5 seconds?

Reference

VTEX also maintains an open reference implementation for the External Seller service under the vtex-apps/external-seller-example GitHub repository (useful as a scaffold; align behavior with the official protocol docs above).

Weekly Installs
61
GitHub Stars
16
First Seen
Mar 17, 2026
Installed on
github-copilot54
claude-code53
codex34
opencode30
gemini-cli30
cursor27