A free Jev API key is the question I get from agency owners who want to sort leads and emails with AI without adding another monthly bill.
The good news is that the cheapest way in right now costs nothing and needs no key at all.
OpenCode Zen runs a model called jev-1.13-free that answered my keyless test call on 5 October 2026 with a cost of "0".
That means you can prove Jev works on a client's real enquiries before you spend a penny or sign anything.
In this playbook I'll show you what Jev is, how to switch it on for free, which jobs it does well for a business, and how I'd turn it into something a client pays for.
Why a business owner should care about Jev at all
Let's start with the money problem, because that's what actually matters here.
Most AI automations quietly spend their budget on tiny decisions rather than big writing jobs.
Is this lead worth a call?
Which inbox folder does this email belong in?
Is this ticket urgent or can it wait until tomorrow?
The old way sends every one of those questions to a big chat model, waits a few seconds, and pays for a full sentence when you only wanted one word.
Multiply that by every email, every form fill and every support ticket, and you're paying premium prices for a yes or a no.
Jev was built to make those small decisions fast and cheap, and the free route lets you test that claim with zero risk.
What Jev is (the plain-English version)
Jev is a decision model from TypeSafe AI that picks an answer from a list you give it, and tells you how likely each answer is.
It launched on 15 September 2026.
You give it a situation, like an email from a prospect.
You give it questions with fixed answers, like "hot or cold" or "billing, technical, sales or other".
It gives you back the answer and the odds, usually in about a second from my desk.
It never writes a word, which is exactly why it's so cheap.
According to the free Jev guide I published on AgentOS.guide, TypeSafe was founded by Diogo Almeida, who previously worked at OpenAI, and it raised a $40 million seed round led by DCVC.
So this isn't a weekend side project that will vanish next month.
How to switch on the free Jev route in five minutes
You don't need a developer for the first test.
You just need to be comfortable pasting one command into a terminal.
Step 1: Pick one real decision from your business
Don't start with a demo question.
Pick a decision you or your team make twenty times a day, like "is this enquiry a real lead?".
Grab ten real examples, with names and personal details stripped out.
Step 2: Send it to the free model
Paste this into your terminal and swap the state for one of your real enquiries.
curl https://opencode.ai/zen/v1/systemone \
-H "Content-Type: application/json" \
-d '{
"model": "jev-1.13-free",
"state": "Hi, we run three dental clinics and need help getting more bookings from Google. What would this cost and when could you start?",
"questions": {
"lead": {
"type": "choice",
"instructions": "How should we treat this enquiry?",
"criteria": {
"hot": "A real buyer with a clear need and intent to start",
"cold": "Curious, vague or not a fit",
"spam": "Spam, sales pitches or bots"
}
},
"reply_today": {
"type": "noul",
"instructions": "Should someone reply to this today?"
}
}
}'
There's no key in that request, and that's deliberate.
The jev-1.13-free model name is what makes it free.
If you drop the "-free" part, Zen rejects the call with a 401 error saying rate-limited models require a workspace.
Step 3: Score the results against your own judgement
Run your ten examples and write down where Jev agreed with you.
Look at the confidence number on the ones it got wrong.
In my experience, the wrong answers tend to cluster at lower confidence, which is the signal you'll use to route those to a human.
🔥 Want the lead-routing setup I use? Inside the AI Profit Boardroom, I walk through how a decision layer like Jev sorts leads and grades enquiries inside the Agent OS. You get step-by-step video tutorials, weekly coaching calls and 3,400+ members building real automations for their businesses. → Get access here
How to use Jev for real business jobs
Here are the jobs where I'd put Jev to work first, because each one is a repeated decision with fixed answers.
| Business job | The question you ask Jev | The answer type | What happens next |
|---|---|---|---|
| Lead grading | Is this enquiry hot, cold or spam? | Choice | Hot leads get a same-day call |
| Inbox sorting | Which folder does this email belong in? | Choice | The email is filed automatically |
| Support routing | Which team should handle this ticket? | Choice | The ticket lands in the right queue |
| Urgency scoring | How urgent is this, from 0 to 4? | Score | Critical issues trigger an alert |
| Reply check | Does this email need a reply today? | Noul | It goes on today's list or waits |
| Task check | Is this task actually finished? | Noul | The job closes or gets flagged |
None of these jobs needs a single sentence written.
That's the sweet spot.
What my own free-model runs showed
I've run these exact kinds of jobs on the free model for my Jev guides.
A 48-email sample inbox went through jev-1.13-free, and 45 of the 48 landed in the right folder in 8.4 seconds.
Twenty-four contact-form messages were graded hot or cold, and all twenty-four landed on the right side.
Those were sample datasets, not a client's live inbox, so test on your own data before you promise anything.
What it cost when I ran it through Vercel
On Vercel AI Gateway, I ran 1,102 real decisions through Jev in one morning.
That used 146,948 input tokens and cost $0.0062 in total, according to the prices on Vercel's model card at the time.
The trick that made the free tier enough was packing ten items into every request, because the free tier capped me at 30 requests a minute.
The Jev free tier options compared for a business
There isn't one official Jev free tier, so here's how the free and near-free options stack up for someone running a business.
| Option | Cost to start | Card needed? | Where I'd use it |
|---|---|---|---|
OpenCode Zen jev-1.13-free |
Free, limited time | No | Proving the idea on anonymised samples |
| jevplayground.com | Free to try, usage limits apply | No | Showing a client what Jev does in a browser |
| Vercel AI Gateway | $5 monthly free credit for eligible models | Yes, to unlock the credit | Small production jobs if Jev is eligible on your account |
OpenCode Zen jev-1.13 |
$0.042 per million input tokens, output free | Yes, via a Zen workspace | Cheap production once the free window closes |
| TypeSafe API direct | $42 per billion input tokens on TypeSafe's site | Yes, a TypeSafe key | Client work where you want the maker's own API |
Vercel's promotional pricing for Jev ended on 25 September 2026, and its model page now lists $0.04 per million input tokens.
Vercel's docs also say that buying credits moves your account onto the paid tier, so if you want the free credit, add the card and buy nothing.
How to turn a free Jev API key route into client income
This is where the business angle really kicks in.
A client doesn't care what model you use.
They care that their leads get answered faster and their team stops wasting hours sorting email.
Sell the outcome, not the model
I'd package Jev as part of a lead-response or inbox-triage service rather than selling "Jev setup".
The pitch is simple: every enquiry gets graded the second it lands, and the hot ones get a reply the same day.
My post on how to start an AI agency covers the offer side in more detail.
Use the free model for the pilot
Run the client's anonymised historical enquiries through the free model during your pitch.
Show them where Jev agreed with their team and where it flagged low confidence for a human.
That's a far stronger sales conversation than a slide deck.
Move production to a keyed route
Once the client signs, move the live workflow to a keyed route like the paid Zen model, Vercel or TypeSafe's own API.
The free model is labelled limited time, and you don't want a client's lead flow depending on a free window.
The running cost is tiny at these prices, so it barely touches your margin.
Pair it with a writing model
Jev decides, and a writing model writes.
So Jev picks which leads get a reply, and a model like Claude drafts the reply for a human to approve.
If you're building out more automations for small businesses, my list of AI automations for small business has the jobs I'd hand off first.
🚀 Want help packaging this for clients? The AI Profit Boardroom is where I share the Agent OS, the 30-day roadmap and the coaching calls where members bring their own client setups to get fixed live. There are 3,400+ members inside, and plenty had never touched AI before joining. → Join the Boardroom here
The limits that could cost you a client
I'd rather you hear these from me than from an angry client.
- Jev's "zero hallucinations" promise is about the shape of the answer, meaning it always picks one of your options, not that the option is correct.
- When I asked Jev to file "Can I bring my dog to the workshop?" with no "other" option, it put it under sales at 98%.
- Jev reads text only, so it can't judge a screenshot or a voice note unless you describe it in words first.
- Jev explains nothing, so when it's wrong you fix the question or the input, not a chain of reasoning.
- The free Zen model may have different data terms, so read the Zen data policy before sending real customer data.
The fix for most of these is the same.
Always include an "other" option, always use the confidence number to send unsure cases to a person, and always strip personal data from free-tier tests.
Who should use the free Jev route
It's ideal for agency owners who want to pilot AI lead grading without adding a new bill.
It's ideal for small businesses drowning in email who want to see what automated sorting looks like first.
It isn't for anyone who wants an AI to write their customer replies, because Jev simply doesn't write.
If you want more ways to earn with tools like this, my list of AI side hustles that work is a good next read.
Related reading
- How to start an AI agency
- AI automation for small business: 7 jobs to hand off
- 12 AI side hustles I'd actually start
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
Also On Our Network
- 🌐 Step-by-step setup for the keyless Jev model
- 🌐 Which free Jev route actually wins: my verdict
- 🌐 Free Jev inside an agentic operating system
- 🌐 How I ran 1,102 Jev decisions on Vercel's free tier
About Julian
I'm Julian Goldie, an AI entrepreneur, SEO expert, and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation, and SEO.
- I have 400,000+ YouTube subscribers who watch my AI tool tests every week.
- I built a 7-figure agency from the ground up.
- I run daily AI training inside the Boardroom.
- I wrote two Amazon best-sellers on SEO and agency growth.
If you want a hand applying this to your own business, you can book a free strategy session.
→ Get my best AI training inside the AI Profit Boardroom
Jev Free Tier And API Key FAQ
Can I get a free Jev API key for business use?
The free route on OpenCode Zen doesn't use a key at all.
You call the jev-1.13-free model at https://opencode.ai/zen/v1/systemone, and it returns answers with a cost of "0".
OpenCode labels it limited time, so use it to prove the idea and move client work to a keyed route.
How much does Jev cost once the free tier runs out?
OpenCode Zen lists the paid jev-1.13 model at $0.042 per million input tokens, with output marked free.
Vercel lists Jev at $0.04 per million input tokens, and TypeSafe's site lists $42 per billion input tokens.
What business jobs is Jev good at?
Jev is good at small, repeated decisions with fixed answers, such as grading leads, sorting emails, routing tickets and scoring urgency.
It can't write replies, so pair it with a writing model for anything that needs words.
Do I need a card for the Jev free tier on Vercel?
When I set it up in September, Vercel returned a verification error until I added a payment method, even though I bought nothing.
Vercel's docs say buying credits moves your account to the paid tier, so add the card but don't purchase credits if you want to stay on the $5 monthly free credit.
Is the free Jev route safe for client data?
Free models can come with different data terms, so read the OpenCode Zen data policy before you send real client data.
For sensitive work I'd use anonymised samples on the free model and switch to a keyed route under the client's agreement for production.
Can Jev be wrong even with zero hallucinations?
Yes, it can.
TypeSafe's zero hallucinations claim means Jev always answers inside the options you give it.
In my testing it filed a dog question under sales at 98% because there was no "other" option, so always add one and use the confidence number.
The bottom line
Jev turns the hundreds of tiny decisions in your business into fast, near-free calls, and the keyless Zen model lets you prove that on real enquiries today.
Pilot it free, sell the outcome, and move live client work to a keyed route before the free window closes.
That's how I'd turn a free Jev API key into actual income.
