For your own tools
Developers
On this page
Read your store from your own tools: your list, your community's members, a course's students, your affiliates and your bookings; and, with a key made to write, add people to your list, label them, and give or take back products. Every plan has it. To be told when something happens instead, use webhooks.
Keys
Make a key in your studio, under API keys. It is shown once, when you make it, and never again: we keep only a fingerprint of it. A store can have 5 keys; each shows when it was last used, and revoking one refuses the very next request made with it. Deleting a store revokes all of its keys and removes its webhook endpoints with it.
A key reads, and changes nothing, unless you check "It can also write" when you make it. That choice is the key's for good: a key made to read stays one that reads, and every key made before writing existed reads only. A key that reads only gets 403 read_only from the addresses that write. Only the store's owner and Admins can make a key that writes.
Send it on every request:
curl https://marktmorgen.com/api/v1/store \
-H "Authorization: Bearer nl_live_…"A key belongs to one store and reads only that store. Keep it on a server, in Zapier or in Make — never in a web page, where anybody who opens the page could copy it. The API sends no CORS headers, so a browser refuses to use one.
Answers, pages and limits
Every answer is JSON. A list comes as data, and a long list comes a page at a time with next: send it back as ?cursor= for the next page, until it is null. A cursor a list did not give is answered 400 invalid_cursor: start again without one. A page holds about 200 rows. Products never has more; your list, your members and a course's students can have a few more on a page, and a page can come back empty before the last one, so keep asking until next is null. Times are ISO 8601, in UTC. Amounts are in the currency's smallest unit — cents for dollars.
To get only some fields of each row, add ?fields= with their names, separated by commas, for example /api/v1/leads?fields=email,agreed. The paging fields beside data always stay. A field no row has is answered 400 unknown_field, with its name, and a name that cannot be a field (a field name is lowercase letters, digits and underscores, starting with a letter, up to 40 characters) or more than 30 names is answered 400 fields. Both are answered before anything is read, so they count as one request. This works on every list above, on Store, on Bookings, on Discount codes and on Sales.
Each key may make 120 requests a minute. Past that, the answer is 429 with Retry-After: 60.
A store's keys together also have a figure for each billing month (the month that starts on the day the store's plan renews, or on the day it had before a switch between monthly and yearly): 500 requests on Store, 2,000 on Studio and 10,000 on Scale. A free trial has 100 for the whole trial, and a store without a paid plan has none. A contact added or a product given counts as 5, and a page of sales as 10, and more when it names products that are read for their titles. Every read except Store counts as more than one when it reads more than one request's worth: a page of products, of your list or of your members, a course's students, your numbers, your discount codes, your affiliates, your bookings and the recent events (/api/v1/events, the examples a tool such as Zapier shows while you set up a step: from 2 to 14 each, by the size of the deliveries it reads, measured before they are read; a webhook message takes at most 62 KB to keep, each text in it at most 256 characters, so 14 is the most for messages made since November 4, 2026, and one kept from before counts at its size), as said with each below; what it will take is taken before it reads. A read whose size is known only once it is read (a page of your list or your members, your numbers from their copy) sets aside the most it may take first, and gives back at once what it did not need. Past the figure, or when a read needs more than are left, the answer is 429 monthly_limit, and its message says when the figure opens again, if it does. A trial or a store without a paid plan also needs one of the month's places for such stores; without one, the answer is 429 no_place. Nothing is billed for requests. A write that does not fit the store's data room (its words and records, by plan) is answered 409 store_full, and nothing is written.
401 {"error": "unauthorized", "message": "Send your API key as: Authorization: Bearer nl_live_…"}
429 {"error": "rate_limited", "message": "At most 120 requests a minute per key. Try again in a minute."}
429 {"error": "monthly_limit", "message": "Your store's API requests for this month are used: 2,000 API requests. This opens again with your next billing month, on …"}
429 {"error": "monthly_limit", "message": "This read takes 6 of your store's API requests, more than your store has left, so nothing was read."}
429 {"error": "monthly_limit", "message": "This read may take up to 14 of your store's API requests, more than your store has left, so nothing was read."}
400 {"error": "unknown_field", "message": "No item in this answer has a field named “nothing_here”."}
400 {"error": "fields", "message": "“bad-name” cannot be a field name. A field name is lowercase letters, digits and underscores, starting with a letter, up to 40 characters; ask for at most 30 fields."}
429 {"error": "no_place", "message": "Stores without a paid plan are full for this month, so this can't run right now. …"}
409 {"error": "store_full", "message": "Your store's data room is full, so nothing was added. …"}
403 {"error": "key_paused", "message": "This key is paused: the store's team now logs in with a second step, and the key was made without one. …"}
409 {"error": "paused_unpaid", "message": "Nothing was read: your plan's renewal payment did not go through, …"}
409 {"error": "paused_unpaid", "message": "Nothing was changed: your plan's renewal payment did not go through, …"}
502 {"error": "unavailable", "message": "… could not be read just now. Try again in a moment."}A missing, mistyped and revoked key all get the same 401.
A key made by a team member who logged in without a second step is paused when the store's owner starts asking the team for one. Every request with it is answered 403 key_paused until whoever made it logs in with a passkey or an authenticator app's code and turns it back on under API keys in the studio. A key made by someone who then leaves the team is paused too, and can only be revoked; one made by someone whose role no longer allows it waits for the store's owner to turn it back on. A paused key is never deleted. The webhook endpoints a key added are sent nothing while it is paused, start again when it is turned back on, and stop for good when it is revoked.
While the store's plan renewal is unpaid, reads answer 409 paused_unpaid (or paused_disputed while a payment for the plan is disputed, and for the period that payment covered if the dispute is decided for the store, while the plan runs), before the request counts toward the month. Adding a lead, adding labels, giving a product and adding a webhook endpoint get the same 409, also before they count; its message starts "Nothing was changed". Removing labels, taking back access and removing a webhook endpoint are not paused: they still work and count as API requests as usual.
Store
GET /api/v1/store Which store the key reads. The first call to make, to check a key works.
{
"data": { "handle": "harbor", "name": "Harbor Kitchen", "currency": "usd" }
}Products
GET /api/v1/products Every product, in the order your store shows them, drafts included.
A full page counts as up to 16 of your API requests, by how many of its products have to be read; a smaller page counts as fewer.
{
"data": [
{
"id": "k3j9x2p1ab",
"title": "The 7-Day Reset",
"price": 700,
"currency": "usd",
"kind": "one_time",
"interval": null,
"interval_count": null,
"hidden": false,
"url": "https://marktmorgen.com/@harbor/p/k3j9x2p1ab"
}
],
"next": null
}kind is one of one_time, free, membership, course or call. interval is set for a membership, and interval_count says how many of them one charge covers: 3 with month is every 3 months.
Leads
GET /api/v1/leads Your list: everybody who asked for something free, bought, or came in with an import — the same rows as the CSV export in your studio. Add ?agreed=true for only those who agreed to hear from you and have not unsubscribed: the ones it is fair to email. If a page holds a contact that could not be read, it leaves that contact out and says how many in unreadable; a page without that field left nobody out.
A page counts as at least 2 of your API requests (one while your store has no list yet), and more when its rows are long, by the bytes it reads and sends: for example, 3 for 200 rows of about 2 KB each. A page needs 14 requests free before it is read, what a full page of the longest rows could take; what it did not use is given back at once. With fewer free, the page is refused and nothing is read. Members count the same way: a page of members needs 8 free before it is read.
{
"data": [
{
"email": "dana@example.com",
"name": null,
"agreed": true,
"agreed_at": "2026-09-12T14:03:22.000Z",
"unsubscribed": false,
"unsubscribed_at": null,
"first_at": "2026-09-12T14:03:22.000Z",
"last_at": "2026-09-20T09:41:10.000Z",
"products": ["Free Pantry Guide", "The 7-Day Reset"],
"product_ids": ["p8w2m4q9zz", "k3j9x2p1ab"],
"tags": [],
"source": "free"
}
],
"next": "1840"
}Members
GET /api/v1/members Everybody who has come into your community, as your own member list shows them. 404 with no_community when the store has none.
{
"data": [
{
"email": "dana@example.com",
"name": "Dana Ortiz",
"handle": "danaortiz",
"listed": true,
"announcements": true,
"muted": false,
"removed": false,
"joined_at": "2026-09-14T18:22:05.000Z",
"last_seen_at": "2026-09-29T21:10:44.000Z"
}
],
"next": null
}Students
GET /api/v1/students?product=<id> Everybody who has opened one course, with how far they got, a page at a time, in no set order. product is the course product's id, from Products, and total is how many students the course has in all.
Changed on November 4, 2026: students now come a page at a time, about 200 a page in no set order, and no longer as the 500 most recently active. If the course changes while you page through it, a student can show up twice, or a page can come back empty before the last one, so the rows you add up may not match total.
Each page reads every student's progress, so a full page of 200 counts as 16 of your API requests (26 when more than 200 students of the course are blocked). A smaller page counts as fewer.
{
"total": 1,
"data": [
{
"email": "dana@example.com",
"lessons_done": 7,
"lessons_total": 12,
"started_at": "2026-09-15T10:00:00.000Z",
"last_seen_at": "2026-09-29T08:15:31.000Z",
"blocked": false
}
],
"next": null
}Numbers
GET /api/v1/stats Your numbers for the last 7, 30 and 90 days, the same ones your studio shows: visitors, checkouts, sales and revenue, each channel from visit to sale, and page views by country and by device.
Read anew, it reads 180 days of counts and, with your Stripe account connected, every page of your sales, renewals, refunds and disputes it may need, so it counts as 40 of your API requests (21 for a store with no sales to read), and more when you have PayPal sales or refunds. A copy of the numbers is then kept for an hour: within that hour, a read comes from the copy and counts only for reading it and its answer, as few as 2 and at most 5. asOf says when the numbers were read, in ISO 8601 (UTC), so a number up to an hour old is never taken for one of this minute.
{
"data": {
"currency": "usd",
"sales": "ok",
"days": {
"last_7_days": {
"visitors": 412,
"page_views": 655,
"checkouts": 31,
"sales": 9,
"revenue": 26100,
"channels": [
{ "channel": "instagram", "visits": 240, "checkouts": 19, "signups": 4, "sales": 6, "revenue": 17400 }
],
"countries": [{ "country": "US", "views": 388 }],
"devices": [{ "device": "mobile", "views": 521 }],
"previous": { "available": true, "visitors": 335, "page_views": 540, "checkouts": 24, "sales": 7, "revenue": 20300 },
"refunded": 2900,
"refunds_complete": true,
"payments": { "complete": true, "gross_receipts": 31100, "after_refunds": 28200 },
"renewals": { "complete": true, "count": 2, "revenue": 5000 },
"payment_channels": {
"complete": true, "from": "2026-10-08", "to": "2026-10-14",
"rows": [{ "channel": "instagram", "gross_receipts": 31100, "share": 1, "identified": true }]
},
"sales_by_country": [{ "country": "US", "sales": 5, "revenue": 14500 }],
"ads": [{ "ad": "reel-b", "channel": "instagram", "campaign": "launch", "sales": 3, "revenue": 8700 }]
},
"last_30_days": { "...": "the same fields" },
"last_90_days": { "...": "the same fields" }
}
},
"asOf": "2026-10-14T16:05:12.000Z"
}Money is in the smallest unit of your store's currency, before Stripe's fee. sales and revenue are null when Stripe could not be read ("sales": "error") or is not connected yet ("none"). A channel of direct is a visit or sale from nowhere the store could name. previous is the same number of days just before, to compare with; its sales and revenue are null when Stripe could not be read that far back in full.
Each period also includes payments and renewals. Gross receipts include initial purchases and renewals. The amount after refunds subtracts refunds recorded in the same period, including refunds of older purchases, before fees and ad spend; it is not profit. Each group has its own completeness flag. Unavailable values are null, while confirmed empty readings are zero. Complete gross receipts remain available when only refunds are unavailable. Initial sales and revenue also remain null when the purchase reading is incomplete, even if the provider answered successfully.
payment_channels lists every channel with positive gross receipts, including renewals, in the stated UTC dates. Each share is a fraction of all recorded receipts (0 to 1). Direct, unrecorded and imported origins have identified: false and remain in the denominator. The list is not capped. An empty list with complete true means no receipts; complete false means the reading is unavailable. Refund availability does not block confirmed receipt channels.
payment_comparison compares gross receipts with the immediately preceding period of equal length for 7 and 30 days, including channels present in only one period. It provides the earlier UTC dates and total, changes in money, fractional growth and changes in channel share. Multiply share changes by 100 for percentage points. Growth is null when the earlier amount is zero. Comparison is unavailable for 90 days or when either period cannot be read completely; complete is false and rows are empty, while confirmed current receipts may remain available.
refund_activity provides refund records and amounts by channel for the stated UTC dates. Its share compares refunded money with gross receipts in the same period, including renewals. It is not the fraction of buyers or purchases refunded: refunds can concern older purchases and the ratio can exceed 1. Share is null when receipts are zero. A complete empty list means no refund activity; incomplete readings have null totals and an empty list. Confirmed gross receipt reports remain available independently.
buyer_channels groups buyers by the channel of their first observed purchase in each 7-, 30- or 90-day window, not their lifetime first purchase. It includes buyers, repeat buyers, orders and observed purchase revenue before refunds and fees, excluding subscription renewals. Thirty-day repeat rates are fractions based only on buyers with 30 full days observed; the rate is null when none are eligible. Unidentified orders are reported separately. Incomplete purchase history makes this group unavailable, with null unidentified-order counts and empty rows. A complete empty group is a confirmed empty reading. No buyer identities are exposed.
At the response root, weekly_buyer_cohorts groups buyers by the UTC Monday of their first observed purchase week. It includes the observation date and 7-, 30-, 60- and 90-day eligible-buyer counts, repeat counts and fractional rates. Each rate uses only buyers with the full observation window; no eligible buyers means null, not zero. History is limited to the purchases read by the platform. Missing or incomplete history makes this group unavailable, while a complete empty reading remains available. Unidentified orders are separate and no identities are included.
Each period includes daily_payments, with UTC dates, gross receipts including renewals, refunds recorded that day and amounts after refunds, before fees and ad spend. Separate payment and refund coverage flags distinguish unavailable readings from real zero amounts. Missing refund coverage leaves confirmed daily receipts available but refunded and after_refunds null. Complete readings include days with zero activity. Refunds of older purchases can produce a negative daily result. Incomplete payment readings have no daily rows.
campaign_results exposes every recorded campaign in each period: visits, paid purchase records, purchase revenue before refunds and fees, and the channels of those payments. Campaign tags aggregate across channels; they do not provide a traffic split by channel. Subscription renewals are excluded. Incomplete purchase readings preserve campaign visits while paid amounts and payment channels remain null. A null channel list means unavailable origin detail; an empty list means zero paid records. Available empty reports are distinct from missing reports.
The ads and sales_by_country lists return at most 20 entries. Their total counts and truncated flags indicate whether more recorded entries exist. The existing completeness flags describe reading coverage, not whether the response contains every entry. Unknown readings have null counts and truncated flags; confirmed empty readings have zero counts and false flags. Do not sum a truncated list as the full period total.
Use /api/v1/stats?details=all to retrieve every recorded ad variant and buyer country in each period. Default requests retain the 20-entry limit. The same authentication and reading-coverage rules apply: requesting all details does not expose incomplete paid readings. Complete full lists have false truncation flags. Ad variants keep their channel and campaign, so the same creative name in different campaigns or channels remains distinct.
attribution compares purchase counts and revenue by the last visit, first visit and the origin buyers reported, using the same channel groups as the studio. These are alternative views of the same purchases; do not add the columns. Unidentified last and first visits remain separate, with survey answers for unidentified last visits. Missing first-visit history leaves its figures null and first_visit_complete false. Incomplete paid readings make the entire comparison unavailable. The configured attribution window is reported when known. Renewals are excluded, and no individual survey answers or buyer identities are exposed.
Each window includes period with UTC start and end dates, the requested length and the number of returned days. Calendar completeness requires the full requested sequence of consecutive valid dates; gaps, duplicates and short histories make it false. Missing or invalid dates are not replaced with invented dates. This describes date coverage, independently of payment coverage. The last day can still be in progress; complete calendar coverage does not mean that today has finished or that financial readings are complete.
previous.available indicates whether the preceding period was read. When it is false, previous visitors, page views, checkouts, sales and revenue are null, rather than invented zeros. Confirmed empty periods retain zero traffic. Availability of previous traffic does not confirm paid coverage: previous paid values still require a complete purchase reading. Current results remain independent of missing previous history. Availability is carried from the underlying traffic reading, so older snapshots without that history do not acquire an artificial zero baseline.
Ask your AI assistant
Marktmorgen is also an MCP server, so an assistant that speaks the Model Context Protocol — Claude, ChatGPT, Cursor and others — can answer questions like "which channel made me money this month?" from your own numbers. Add it as a remote server at https://marktmorgen.com/api/mcp, with the header Authorization: Bearer nl_live_… and one of your keys.
It has twenty-four tools. Three prepare a change: a new description, title or short summary for a product, each kept as a proposal you read beside the current one and confirm under What the AI proposes in your studio, with undo; nothing changes until you confirm, and they need a key you allowed to prepare changes. The other twenty-one read only: your numbers (the same as /api/v1/stats), your products, revenue by channel read three ways, the funnel from visit to sale, each product's numbers and what buyers buy next, your buyers in numbers, what is worth knowing, your memberships (MRR, churn, why members left, month by month), your emails and sequences with what they sold, your affiliates by their codes, a product's description and its sales page block by block (testimonials counted, never quoted), a course's outline and each lesson's text, quiz and task (files named, never located), how a product's page test is going, how far down its sales page visitors read and its earlier versions, the sections you saved to reuse, which products can go in a bundle and why the others cannot, how your imports from another platform went, in counts, and which store the key belongs to. Most of the ones about a stretch of time take days: 7, 30 or 90. It changes nothing, and it does not read anybody's name or address, buyers', members', subscribers' or affiliates': what an assistant reads is kept by whoever runs the assistant, and the people on your list did not agree to that. The endpoints above are where your own tools read them.
Each tool counts as many of your API requests as the address that reads the same: your numbers and the tools about a stretch of days as /api/v1/stats, read anew or from the copy kept an hour (the tools about a stretch of days keep their own copy, with the products they name, and count those products when read anew), each answer with its asOf; each product's numbers also count what buyers bought next, by its size; your products as /api/v1/products and your affiliates as /api/v1/affiliates. Your memberships count as about 3, or 17 when they were not read in the last ten minutes, and more for the names of the products they name; your emails count by how many emails and sequences you have, and more when what they sold was not read in the last ten minutes. A product's description, its sales page, a course's outline or one lesson, a page test and how far a page is read each count as a few, by the most they could read; a page's earlier versions and your saved sections count more as more are kept. A page of products for a bundle and your imports each count as a few, by the most they could read. Preparing a description, a title or a short summary counts as a few, more as more proposals wait or are kept. Which store the key belongs to counts as one.
Affiliates
GET /api/v1/affiliates Every affiliate, with what they earned, were paid and are owed — the numbers on your affiliates page.
{
"currency": "usd",
"refunds_checked": true,
"data": [
{
"id": "0123456789ab",
"email": "sara@example.com",
"code": "sara",
"status": "approved",
"applied_at": "2026-09-02T16:40:12.000Z",
"clicks": 214,
"sales": 9,
"earned": 18900,
"paid": 12000,
"owed": 6900,
"payable": 4200,
"waiting": 2700
}
]
}It reads your whole program, so it counts as at least 3 of your API requests, and more as your program grows: for example, 17 for 100 affiliates with 1,000 sales and your Stripe account connected. The count is made before anything is read.
payable is what can be paid today; waiting is still inside the wait you set before a sale can be paid. When refunds_checked is false, Stripe could not be asked about refunds just now and the amounts may be high: read again before paying anyone from them.
Bookings
GET /api/v1/bookings Paid calls and seats in live sessions, each at the time it is booked for now — a booking its buyer moved is at its new time.
With your Stripe account connected, it reads up to 500 bookings from it, so it counts as 6 of your API requests, however many you have. Without Stripe connected, it counts as one.
{
"data": [
{
"id": "cs_live_a1B2c3…",
"product_id": "c9v7n2k4ww",
"title": "30-minute strategy call",
"email": "dana@example.com",
"name": "Dana Ortiz",
"starts_at": "2026-10-03T15:00:00.000Z",
"ends_at": "2026-10-03T15:30:00.000Z",
"buyer_time_zone": "America/New_York",
"moved": 0,
"amount": 9900,
"currency": "usd",
"answers": [{ "label": "What should we cover?", "value": "Pricing my first course" }]
}
]
}Writing: your list
POST /api/v1/leads Puts somebody on your list, as someone who agreed to hear from you. Send agreed: true, your word that they did: nobody is added without it, and nobody who unsubscribed from your list is ever added back, whatever is sent. Somebody new starts the sequences that start when somebody joins your list; a label they come with starts the sequences that label starts.
curl -X POST https://marktmorgen.com/api/v1/leads \
-H "Authorization: Bearer nl_live_…" \
-H "Content-Type: application/json" \
-d '{"email": "dana@example.com", "name": "Dana Ortiz", "tags": ["webinar"], "agreed": true}'
200 {"data": {"email": "dana@example.com", "status": "added"}}
200 {"data": {"email": "dana@example.com", "status": "already"}}
400 {"error": "consent", "message": "Send agreed: true, …"}
409 {"error": "unsubscribed", "message": "This person unsubscribed from your list, so they are not added back."}POST /api/v1/leads/labels Adds and takes away labels of somebody already on your list. A label added starts the sequences it starts in your studio.
{"email": "dana@example.com", "add": ["vip"], "remove": ["webinar"]}
200 {"data": {"email": "dana@example.com", "labels": ["vip"]}}
404 {"error": "not_on_list", "message": "Nobody with this address is on your list."}A store may add 2,000 new people a day through the API; past that, 429 daily_limit until the next day.
Discount codes
GET /api/v1/codes Your discount codes, read from your own Stripe account as your studio lists them: what each takes off, how many times it was used, its limit, its smallest order and whether it is live. One answer holds your 20 newest codes; has_more is true when you have older ones too, which you can see in your Stripe dashboard.
Reading them asks your Stripe account, so it counts as 2 of your API requests; making a code counts as 3 and switching one off as 3, taken before Stripe is asked.
POST /api/v1/codes Makes a code on your own Stripe account, checked as in your studio. Send {"code": "LAUNCH20", "percent": 20}, or {"code": "TENOFF", "kind": "amount", "amount": "10"} in your store's currency; uses and minimum are optional. A code that cannot be made says why, as error and message; the same letters twice give 409 taken.
DELETE /api/v1/codes?id=promo_… Switches one of your store's codes off. Stripe never deletes a code that may be on a receipt, so it stays in the list, switched off. A code set to start later loses its start date too, so it never comes back on by itself; while that start is being handled the answer is 409 busy, and trying again a moment later works. If it was switched off but your codes could not be read back from Stripe, the answer has stopped with its id, listed: false and data: null: read them again with GET. Making and switching off need a key that writes; all three answer 409 not_selling while your store is not taking payments.
Writing: giving a product
POST /api/v1/access Gives a product to an address with no payment, as an import of past buyers does: a file, a link, a course, a podcast or a bundle, by its id or its exact name. They open it from your store's list of purchases with that address, where it shows as given by your store, not as a sale; it is not counted in your numbers, sent to webhooks or credited to an affiliate. Send notify: true and they get one email from your store's name saying so, in your store's language, while your plan is running. Somebody deleted from your store is never given anything again or emailed.
{"email": "dana@example.com", "product": "k3j9x2p1ab", "notify": true}
200 {"data": {"email": "dana@example.com", "product": "k3j9x2p1ab", "status": "given", "emailed": true}}
200 {"data": {"email": "dana@example.com", "product": "k3j9x2p1ab", "status": "already", "emailed": false}}
409 {"error": "cannot_give", "message": "Memberships, calls and products with several prices cannot be given this way."}
409 {"error": "deleted", "message": "This person was deleted from your store, so nothing is given to them and they are not emailed."}DELETE /api/v1/access Takes back a product the API gave, with the same body. Only what the API gave: a sale, an import, a gift or a purchase paid with PayPal is not the API's to take, and gets 404 not_given_here.
{"email": "dana@example.com", "product": "k3j9x2p1ab"}
200 {"data": {"email": "dana@example.com", "product": "k3j9x2p1ab", "status": "taken_back"}}A store may give 1,000 products a day through the API. A body is JSON of up to 4,000 bytes.
Sales and refunds
GET /api/v1/sales Your store's paid checkouts, newest first, read from your own Stripe account as your studio's list of sales reads them, in the same shape as the sale.completed webhook. Up to 100 checkouts are read per call, and each call counts as 10 API requests, because it waits on Stripe. A page can be empty and still have a next, when those checkouts were not this store's. A checkout that only saved a card is not a sale and is not listed. What a buyer typed at checkout is not included. 409 with not_selling while the store is not taking payments.
{
"data": [
{
"checkout_session": "cs_live_a1B2c3",
"kind": "one_time",
"product": { "id": "k3j9x2p1ab", "title": "The 7-Day Reset" },
"option": null,
"bundle_items": [{ "id": "q8w2e4r6ty", "title": "Reset Workbook" }],
"order_bump": { "id": "m5n7b9v1cx", "title": "Meal Planner", "bundle_items": null },
"order_bumps": [{ "id": "m5n7b9v1cx", "title": "Meal Planner", "bundle_items": null }],
"amount_cents": 5800,
"discount_cents": 0,
"currency": "usd",
"buyer": { "email": "dana@example.com", "name": "Dana Ortiz" },
"opened_at": "2026-09-20T09:38:52.000Z",
"paid_at": "2026-09-20T09:41:10.000Z",
"refunded_cents": 0,
"disputed": false,
"trial": false,
"source": { "utm_source": "newsletter", "utm_medium": "email", "utm_campaign": null, "utm_content": null, "site": null }
}
],
"next": "cs_live_a1B2c3"
}amount_cents is what the buyer paid at checkout, products checked at checkout included, in the currency's smallest unit, as Stripe counts it: cents for dollars, but whole yen for jpy and whole units for the other currencies Stripe has without decimals, such as krw. So 4900 is $49.00 in usd and ¥4,900 in jpy. discount_cents is in the same unit, and is null if Stripe gave no discount details.
paid_at is when the buyer paid: the charge, or for a membership or payment plan, when its first invoice was paid (with a free trial, that invoice is for $0, so this is when the trial began). It is not when the checkout page was opened. It is null when nothing was charged at checkout, such as a free order. currency is always the three-letter code Stripe gave: if Stripe did not give one for a sale, the answer is 502 unavailable, never a guessed usd.
Products checked at checkout are in order_bumps, in the order of their boxes, and the first is also in order_bump; a bundle's products are in bundle_items, which is null when there is no bundle. A product's title is null when the product is no longer in your store, or, on a page naming more than 200 different products, for those past the 200th. A one-click extra taken after the checkout (an upsell) is its own payment, not a checkout, so it is not on this list: the sale.completed webhook with "kind": "upsell" tells you of each one.
refunded_cents and disputed are not in the webhook, which is sent when the sale is made. For a one-time payment they are read from its charge when you call: how much of it was refunded so far, in the same unit as amount_cents, and whether the buyer disputed it. For a membership or payment plan, and when nothing was charged, both are null: not read here, which is not the same as none. Add the refund.issued and dispute.opened webhooks in your studio to be told of each one, or read them in Stripe's own API with your own Stripe key.
The list is in the order Stripe keeps checkouts: newest first by opened_at, when the checkout page was opened, not by paid_at. A buyer can open a checkout and pay later, and some payment methods, such as bank debits, settle days after that, so a sale can appear below sales you already read. To keep a copy in step, read again from the top each time and go on until a sale's opened_at is earlier than your last read, less a margin: at least 24 hours, and a week or more if your store takes payment methods that settle later. Use checkout_session to skip a sale you already have.
Texts are cut to a fixed length, so a page stays small: a product's title from the checkout to 480 characters, the buyer's name to 150, each utm tag to 40 and the site to 300, with control characters left out. Lengths are counted in UTF-16 code units, as JavaScript counts them, so an emoji counts as 2. An email address longer than 254 characters is not a real one, so it is null rather than cut, and so is one holding a control character. product.id and option are null when the id stored on the checkout is not shaped like an id (letters, digits, hyphens and underscores, up to 64).
A cursor this list did not give, or sent twice, is answered 400 invalid_cursor: start again without one. If Stripe is busy with your account, the answer is 503 with Retry-After. If Stripe's answer cannot be read whole, the answer is 502 unavailable, never a page with guessed amounts.
Checkout on your own website
To sell from your own site, no key is needed or allowed in a web page: the checkout script opens a product's card or your sign-up form over your page, by a link that names only your store and the product, and tells your page and your webhooks what happened.