Enable in-context shopping on AI agentsPrivate preview

To configure your platform and connected accounts for in-context agentic selling, provide the following details:

• Platform Stripe account ID.
• Connected account IDs. Provide a list of connected account IDs to enable for in-context agentic selling. You can send a spreadsheet or a plain text file with one account ID per row.

Prepare your product catalog

Create a CSV file that conforms to the Stripe product catalog specification for each connected account. Stripe shares the specification separately.

Upload the catalog data to Stripe

Deliver the feed through the Stripe APIs. You can send updates every 15 minutes.

Use the sandbox to validate parsing, field mappings, and data quality before enabling live updates.

First, upload your product catalog CSV using the Files API. A successful request returns a File object, which includes the id.

• Specify data_management_manual_upload as purpose.
• Make sure the MIME type matches the file format. Acceptable formats include CSV and TSV, where each row represents one product or variant.
• The maximum file size is 200 MB.

curl https://files.stripe.com/v1/files \
  -u 
sk_test_BQokikJOvBiI2HlWgH4olfQ2: \
  -H "Stripe-Account: {{CONNECTED_ACCOUNT_ID}}" \
  -F purpose=data_management_manual_upload \
  -F file="@/path/to/your/file.csv;type=text/csv"

Then, use the Data Management API to create an ImportSet. This call starts catalog processing and makes the data available in the Dashboard. Include the following:

• The file id returned
• The preview header (for example,Stripe-Version: 2025-09-30.clover;udap_beta=v1)

curl https://api.stripe.com/v1/data_management/import_sets \
  -H "Stripe-Version: 2025-09-30.clover;udap_beta=v1" \
  -H "Stripe-Account: {{CONNECTED_ACCOUNT_ID}}" \
  -u 
sk_test_BQokikJOvBiI2HlWgH4olfQ2: \
  -d file={{FILE_ID}} \
  --data-urlencode standard_data_format="product_catalog_feed"

Monitor feed status

Stripe processes the catalog data, validating and cleaning it before indexing in a format Stripe validates and cleans the catalog data, indexes it, and converts it to a format for AI agents. Monitor indexing progress using the status field on the import set. The status can be pending, failed, succeeded, succeeded_with_errors, pending_archive, or archived.

curl https://api.stripe.com/v1/data_management/import_sets/{{IMPORT_SET_ID}} \
  -u 
sk_test_BQokikJOvBiI2HlWgH4olfQ2: \
  -H "Stripe-Account: {{CONNECTED_ACCOUNT_ID}}"

The response includes the status and any errors:

{
  "id": "impset_7MabcdZ8b617780e5145413",
  "object": "data_management.import_set",
  "created": 1643992696,
  "livemode": true,
  "result": {
    "errors": {
      "file": "file_234923sIENc",
      "row_count": 30
    },
    "rows_processed": 120,
    "successes": {
      "row_count": 90
    }
  },
  "status": "succeeded_with_errors"
}

If your import status is succeeded_with_errors, you can download the error report:

1. Find the result.errors.file field in the response.
2. Use the Files API to retrieve the error file by its ID.
3. The downloaded CSV contains your original data with a leading column named stripe_error_message that describes why each row failed.

curl https://files.stripe.com/v1/files/{{ERROR_FILE_ID}}/contents \
  -u 
sk_test_BQokikJOvBiI2HlWgH4olfQ2:

The API returns a CSV file containing only the rows that failed, with a stripe_error_message column describing each error.

Listen to Stripe webhooks to monitor orders made on AI chat agents.

When an order is confirmed, Stripe emits webhook events that your server can handle to run fulfillment logic. Set up an endpoint on your server to accept, process, and acknowledge these events. See the webhooks guide for step-by-step instructions on integrating with and testing Stripe webhooks.

Stripe emits checkout.session.completed and payment_intent.succeeded. If your fulfillment logic already handles these events, you don’t need additional integration changes. You can customize your fulfillment logic for in-context agentic selling (for example, by noting in your order confirmation email that the checkout occurred through an agent).

const stripe = require('stripe')(process.env.STRIPE_SECRET_KEY);

// Use the secret provided by Stripe CLI for local testing
// or your webhook endpoint's secret
const endpointSecret = 'whsec_...';

app.post('/webhook', (request, response) => {
  const sig = request.headers['stripe-signature'];
  let event;

  try {
    event = stripe.webhooks.constructEvent(request.body, sig, endpointSecret);
  } catch (err) {
    response.status(400).send(`Webhook Error: ${err.message}`);
    return;
  }

  if (event.type === 'checkout.session.completed') {
    const session = event.data.object;
    // Fulfill the order using the session data
    fulfillCheckout(session.id);
  }

  response.status(200).send();
});

You can test your integration directly from the Dashboard in a sandbox:

1. Go to the Agentic Commerce page, then click View catalog.
2. Hover over the product you want to test, then click Test in Workbench.