Powered by the Jev model via the official TypeSafe API
Understand customer messages with Jev AI
Find out what a message is about, how urgent it sounds, and how the customer feels. Paste your text — we’ve prepared the questions.
Jev analyzes the text you provide and returns labels, probabilities, and scores.
Try it — free to start
Pick what you want to understand, look at an example, then analyze your own text. New visitors get free analyses, one time.
New · free beta
Review Analyzer
Upload app reviews and get a clear list of what to check first — each review gets a short label and one next step, plus a labeled CSV download. The first few reviews per visitor are free; no account needed, or sign in to analyze more with a Starter Pack.
What Jev can help with
Route support messages
See what a ticket is about, how urgent it sounds, and how frustrated the customer is — so the right team looks at it first.
Read reviews at a glance
Turn long reviews into sentiment, satisfaction, and a recommendation signal you can compare across many reviews.
Spot sales interest
See what an inbound message is asking for and whether it expresses buying interest, so fast replies go to the best leads.
What to expect
- Labels, probabilities, and scores — plain answers, not JSON you have to decode.
- Probabilities are the model's estimates, not guarantees. We show raw numbers next to every conclusion.
- You stay in control: nothing is sent, replied to, or assigned automatically.
- New visitors get 3 free analyses. Browsing the examples always costs nothing.
Pricing
Starter Pack · US$4.99 · 300 credits — a one-time purchase, enough for 100 review analyses. Paid credits never expire; no subscription.
Explore the Jev ecosystem
20 open-source projects already build on Jev — browser agents, coding tools, MCP integrations, trading bots, data pipelines and games. Browse the directory and see what typed judgments can do.
How the technology works
Jev is a structured-decision model: instead of writing paragraphs, it answers typed questions about your text. The playground turns each answer into the readable conclusions you saw above. Under the hood there are three primitives:
Yes/No probability
the API calls it “Noul”
Ask a yes/no question about your text and get a probability between 0 and 1 — for example, “Is this ticket urgent? → 93%”.
“Is this ticket urgent?” → 0.93
Classification
the API calls it “Choice”
Sort text into one category — billing, complaint, purchase inquiry — and see the probability for every option, not just the winner.
“Category?” → billing 0.88 · technical 0.12
Scoring
the API calls it “Score”
Rate text on a scale, like customer frustration from calm to furious. Scores can land between levels and include the model's own confidence.
“Frustration?” → 2.4 / 4 · confidence 0.9
Questions are fixed on our server per task — you never have to write or tune them yourself.
Frequently asked questions
What is this site?+
ai-jev.net is a free playground for Jev, the structured-decision AI model served through the official TypeSafe API. You paste text, and instead of generated prose you get labels, probabilities and scores you can act on directly.
Is it really free? Do I need an account?+
Yes, it's free and there is no sign-up. Your first visit grants a one-time allowance of 10 free credits, and each analysis costs 3 credits — enough for 3 full analyses. Browsing the built-in examples never uses credits. The grant does not refresh daily; paid credit packs may arrive later, and nothing is charged today.
What kinds of text can I analyze?+
The playground offers three ready-made tasks: understand a customer issue, understand a product review, and understand a sales message. Each checks up to three fixed questions written for that task — you can see exactly what will be checked before you analyze anything.
What happens to the text I submit?+
Your text is sent over HTTPS to our server and forwarded to the TypeSafe API, which performs the analysis. We do not store your text after the response returns, and we don't put it in logs. TypeSafe's public documentation states that models are not trained on customer requests or responses. See our Privacy Policy for details.
How accurate are the results?+
Results are probabilistic model estimates, not guarantees. A 90% urgency probability means the model believes urgency is very likely — not that it is 90% certain to be right in every case. When the signals are close, we say so (“Mixed signals”) instead of picking a winner, and raw numbers are always available under “View analysis details”.
Why did my analysis say the daily budget is used up?+
This site funds all free analyses from a shared daily and monthly budget. When other visitors have used up a day's budget, real analyses pause until it resets (the message shows the exact reset time in UTC). Credits are never charged for a request that was paused this way.
Are you affiliated with TypeSafe?+
No. ai-jev.net is an independent tool that calls the official TypeSafe API. We are not affiliated with, endorsed by, or sponsored by TypeSafe, and 'Jev' and 'TypeSafe' are trademarks of their respective owners.
Which model powers the analyses?+
Analyses run on the Jev model through the TypeSafe API. The exact model version is shown with every result (for example jev-1.13.0, currently behind the jev-latest alias).
See it in action
New visitors get 3 free analyses — no account, no card. Sign in for more with a one-time Starter Pack (US$4.99).
Start analyzing →