TypeSafe serves Jev from one API, and its own docs are the primary source. But search for Jev and a second tier shows up first: independent sites that wrap the same model with their own playgrounds, instant keys, and prepaid credits. They are useful, and they are not TypeSafe. Knowing which is which is the difference between a quick experiment and handing a third party a key that can spend money.
Why a second tier exists
Jev is a hosted model behind a single endpoint, so anyone can resell access or build a front end on it. That has produced a cluster of community hosts, each with its own domain and its own copy of the model, from jevtypesafeai.com and jevtypesafe.org to jevai.org and the jevaimodel family of playgrounds. Their pitches are similar: skip the official signup, get a key immediately, and try the three primitives in a browser.
The disclaimer is the tell
The reliable signal is not the styling, which often mimics the official site, but the footer. The unofficial hosts say so in their own words. jevtypesafeai.com is "operated by CODEFASHION TECH LTD" and "not affiliated with or endorsed by TypeSafe AI". jev-ai.dev calls itself "an independent, unofficial calculator and guide". typesafeai.org is "an independent publication", and jev-tutorial.org is "independent community tutorials". When you see that sentence, treat everything above it as a third party's product.
What the hosted keys mean
The most consequential difference is where the key lives. Some hosts issue their own keys — jevtypesafe.org advertises instant jv_live_ keys against its own /api/v1/decide endpoint and prepaid packs from $5 — so your requests pass through their servers before they reach TypeSafe. Other playgrounds run against the site's own server key, so you never paste one, but then you cannot see the account the traffic is billed to. Either way the data path is longer than calling TypeSafe directly, and the privacy terms in force are the host's, not TypeSafe's.
Where the free material is genuinely useful
Not every unofficial site is a host. Some are explainers. jev-tutorial.org sequences a multi-chapter course from the first call to confidence policy; typesafeai.org collects cases, tools, and guides; jev-ai.dev is mostly a cost calculator that sets Jev's input price beside GPT-5.6 and Claude tables; jev-agent.com wraps a set of open-source models behind the same question shape. These are reading material, and the useful ones cite the official docs rather than replace them.
What to use the official surfaces for
Anything that touches production belongs on TypeSafe's own surfaces: docs.typesafe.ai for the API contract and the primitive limits, the Console for a key tied to your account, and evals.typesafe.ai for the published workflows. The unofficial sites are fine for a first look and for reading, and a poor place to leave a key that can spend your money.
A checklist before you paste a key
- Does the footer say it is unaffiliated? Then its privacy policy governs the request.
- Does it issue its own key, or use its own server key? Both add a party to the path.
- Does it charge credits or a subscription? Compare against TypeSafe's own $0.042 per million input tokens before prepaying.
- Does it cite the docs, or restate them with no source?
- Can you get the same result from the official playground with your own key? If so, prefer that.
The pattern is not unique to Jev: a hosted model behind a single endpoint attracts resellers and front ends within days. The primitives look the same on every one of these sites, and so does the advice — read with them, but keep the key on the official account.