The Faxvio data API: four fax datasets, keyless and free to reuse

Where these numbers come from

Everything here is derived at build time from the app itself rather than typed into a file. The destination list is read from the Faxvio app's own server configuration — the allow-list the app checks a number against before a page is charged — so the API publishes the rule rather than a description of it. The prices are the four App Store product records, whose identifiers are immutable, so a pack size cannot silently change underneath a price. The routing rules and the transmission-time constants are the same modules the free tools on this site run on, which are in turn a port of the app's own engine, checked against the app's test vectors at every build.

That matters for a reader deciding whether to trust the file. A marketing page and a data file usually drift apart because they are maintained separately; here they are the same source rendered twice, so a disagreement between the JSON and the page would be a build failure rather than something a reader has to notice. The current snapshot is list version 20260902, read 2026-09-06, and both dates are printed in every payload and in the footer of every page.

What it costs to use: nothing, with one condition

There is no key to request. Every endpoint is a static file behind a CDN, served with an open cross-origin header, so browser JavaScript is a first-class consumer and a script that runs once a month is as welcome as one that runs once. The single condition is attribution: the data is published under CC BY 4.0, and reuse asks for a link back to the endpoint or to the page a figure came from. A dataset nobody is sure they may reuse does not get reused, and a link back is the only payment this arrangement collects.

What each field means, and what readers get wrong

A JSON schema can say that a field is a boolean. It cannot say which of two plausible questions that boolean answers, and on this data almost every mistake is that kind of mistake. The table below is the part of the documentation worth reading before writing any code against these endpoints; the same notes are carried inside the payloads, so a file that travels away from this page keeps them.

A worked lookup, end to end

Take a number printed on a letterhead as 06698 1234 with the note "Italy". Strip the formatting, notice that Italy is not in the trunk-zero list, and keep the leading zero: the international form is +39 06698 1234. Now resolve the territory. The dial code +39 appears in the shared-dial-code table with one entry, the prefix 06698, which belongs to Vatican City — so this is not an Italian number at all, and a lookup that stripped the zero first could never have matched it. Vatican City is in the destination list, so the fax can be sent; it carries no special prefixes, so four pages cost four pages.

The same procedure with a North American number turns on the area code instead. In +1 876 555 0188 the first three digits of the national number are 876, which the area-code exception table maps to Jamaica rather than to the United States. Jamaica is not in the destination list, so the app refuses the number before charging anything — and an integration that had assumed +1 meant the United States would have quoted a price for a fax that will never be sent. Both rules live in the routing block of the country endpoint for exactly this reason.

Prices, pages and the doubling rule

Pricing is simple arithmetic with one trap in it. A page costs between $0.18 and $0.36 depending on which of the four packs it was bought in; packs do not expire and there is no subscription. What changes the total is the destination prefix: toll-free, shared-cost and premium numbers cost the carrier more to reach, so each page sent to one uses two. The prefixes are listed per country in the destination endpoint in plain English, and per dial code in the routing block as raw prefixes for code to match against. A fax to an agency's 800 line therefore costs twice what the same fax to its local number would, which is worth saying out loud in any answer that quotes a price.

Two limits bound everything: at most 50 pages in one fax, and at most 19 MB in one document. They are in the payload rather than in prose here because a consumer needs to check them before quoting, not after.

Reading the site itself, without parsing HTML

Every page on this site has a markdown twin. Append index.md to any path, or send an Accept: text/markdown header to the ordinary URL and the same address returns markdown. The twins are generated from the rendered pages at build time, so they cannot fall behind the HTML, and they run about fifteen per cent of its size — the difference between an agent reading a page whole and truncating it halfway through a table.

Three further files complete the picture. /llms.txt is an index of the site with one line per page and a section naming every machine entry point. /llms-full.txt is the full text of every page in one document. /auth.md documents the authentication situation, which is that there is none and none is possible — a statement worth publishing precisely because an agent otherwise has to discover it by guessing.

Tools, for an assistant rather than a program

The same data is exposed as five read-only tools over the Model Context Protocol at /mcp: format a fax number, check a destination, list the destinations, estimate a cost and estimate a duration. Each one returns the figure together with the page it came from, so an assistant that used a tool still has something to cite. An A2A endpoint at /a2a answers the same questions in prose for agents that speak that protocol instead.

What none of them can do is act. There is no tool that sends a fax, spends a page, creates an account or reads anything about a person, because none of those has a server-side interface at all — sending and paying happen inside the app on the user's own device, authenticated by their App Store account. That is why there is nothing here to authenticate and nothing to abuse, and it is why the OAuth metadata this site publishes declares an empty list rather than inventing an authorization server. Faxvio also does not receive faxes, so an agent asked for an inbound number has no answer to give here.

Caching, freshness and fair use

Cache these files. They are static, they sit behind a CDN with a one-hour edge lifetime, and they change only when the app's configuration or the App Store products change. If you need to decide whether to refetch, ask status.json instead: it is under a kilobyte and carries the row counts and the dates the larger payloads were derived from. A consumer that polls the small file and refetches the large ones on a change costs this site almost nothing and never serves a stale price.

There is no enforced rate limit. The reasonable shape of use is a scheduled fetch rather than a request per page view, and a script that hammers a static file gains nothing over one that reads it once and keeps it.

Bottom line

faxvio.com publishes four fax datasets as keyless, CORS-open JSON: 37 destinations an app will dial, 250 territories with the routing rules that resolve a typed number to one of them, four page packs from $0.18 to $0.36 a page, and the ITU-T model behind fax duration. Everything is derived from the app at build time, documented by an OpenAPI 3.1 file at openapi.json, and licensed CC BY 4.0 with a link back. The two fields to read carefully are sendable, which is the app's allow-list rather than the carrier's coverage, and special_prefixes, which doubles the page count rather than the price of a page.