Pro

Not open yet. Everything is free; a paid plan only raises limits.

There is nothing behind a paywall. Every plan has every feature. No endpoint, no field and no dataset is reserved for a paying customer — a higher plan raises how much you may use, never what you may reach. The data is CC0 or a US federal work: it is already yours, free, and you may redistribute it commercially without asking. Charging for access to public-domain records would be charging rent on something that is not mine.

What you are actually paying for

Capacity on a shared server. This runs on one small VPS alongside five other APIs, on two vCPUs. Rate limits exist so one caller cannot take it down for everyone — the scarce resource is my hardware, not the data.

The limits, and what each one means

Limit	What it caps
`requestsPerMinute`	Requests in a rolling minute — by IP when anonymous, by key otherwise.
`requestsPerDay`	Requests in a rolling day, same identity.
`maxPageSize`	The largest `?limit=` a search will accept. Bigger pages mean fewer round trips.
`bulkDownloadsPerDay`	Full static-dataset pulls (`/api/v1/*.json`). `producers.json` alone is 18 MB, so this is bandwidth, not access. The per-record API is counted separately.

Every limit listed here is enforced by the code. If a limit is advertised, it is real — one that was not (burst) was removed rather than left in the brochure.

If you are doing public-interest work, do not pay either

Journalists, academic researchers, nonprofits, civic-tech and transparency projects and public bodies get Pro-level limits and the full history, free, by application. Message me on LinkedIn with a sentence about what you are looking into. That is the whole process.

Before you pay me anything

If you need more than any plan offers, do not buy one. The complete datasets are published at /wine/api/v1/ and the server is MIT licensed. Run your own copy and you have no limits at all. A rate limit is a property of my hardware, not a property of the data, and it would be dishonest to pretend otherwise.

A plan makes sense when you would rather not operate it yourself — and Business adds a response-time commitment and a named contact, which is my time rather than my bandwidth.

What this API does not do

It does not serve TTB COLA label records — the label approvals themselves. The Public COLA Registry publishes no bulk feed and no API, and this service does not scrape it. What is served here is the reference data a label points at. If you need label records, that is a different product and this is not it.

How billing will work

Through PayPal Subscriptions. No card details and no PayPal credential will ever reach this service — it will receive a signature-verified webhook that issues or revokes an API key, and nothing else. There is no payment code in this repository today, by design.

Pro will run on its own domain against this same API. Your existing integration will not need to move, and the free plan will keep working exactly as it does now.

Want it?

It will be built when someone actually hits the ceiling. If you are close, say so — and tell me which limit you are hitting, because that is what sets the shape of the plans.

Tell me which limit you are hitting See the machine-readable plans

Price

Plan	Monthly	Annual	Req/min	Req/day	Page	Bulk/day	History
Free (anonymous)	$0	—	60	5,000	100	50	30 days
Free + key	$0	—	300	50,000	100	500	30 days
Public interest	$0	—	1,200	500,000	1,000	5,000	full
Pro	$19	$190 2 months free	1,200	500,000	1,000	5,000	full
Business	$49	$490	6,000	5,000,000	5,000	50,000	full

Pro and Business are not open yet. Free, Free+key and Public interest are live now.

How that compares, honestly

The closest competitor is COLA Cloud, at $39 and $99 a month (verified 19 September 2026). A straight comparison, including the parts that do not flatter this service:

	COLA Cloud Pro — $99	This service, Pro — $19
Rate limit	120 req/min	1,200 req/min
Volume	1M list records/mo	~15M requests/mo
COLA label records	~2.7M, back to 2005	none
Label images	yes	none
AI enrichment, OCR, barcodes	yes	none
Point-in-time history	no — an `is_active` flag, overwritten	yes
Wine reference depth	AVA list	AVA nesting, 27 CFR variety names + aliases, code tables

The part that does not flatter us. COLA Cloud's free open-data layer already publishes a broader TTB permittee file (~114,000 rows, refreshed daily) and an AVA list, both CC0. So the current-state reference data here is not a differentiator, and it is not what a paid plan buys. What a paid plan buys is request headroom and the retained history.

That is why Pro is $19 and not $39: on breadth of data this service is a subset of what a competitor gives away for nothing, so pricing at or above their entry tier would be charging more for less. $19 also sits inside the band small public-data APIs actually occupy — CourtListener $10, Protomaps $14, Stadia Maps $20.

And the ceiling on all of it: every dataset here is public domain, published in full as static files, and the server is MIT licensed. If a plan looks expensive, self-host and pay nothing. That is a deliberate cap on what this can ever charge.