Get a Jev API key on this host

How to get a Jev API key — full flow

1. Open a hosted tool anonymously and run a few real requests (no signup) so you can see choice / score / noul answers before writing client code. Pick a triage or moderation template close to your domain so the sample state looks like production traffic.
2. Sign in on Account. Registered accounts receive a small monthly grant of free input capacity for wiring and smoke tests — enough to validate auth headers and JSON shape, not to run production traffic or overnight load tests.
3. Buy a prepaid pack on Pricing (from five dollars). Larger packs lower the retail dollars-per-million rate and raise rate limits. Capacity does not expire while the account stays active, so you can stage validation on a small pack and top up when volume appears.
4. On Account, create an API secret. Copy the jv_live_ value once, store it as an environment variable such as JEV_API_KEY, and never commit it to git or ship it inside a mobile or browser bundle.
5. Send POST /api/v1/decide (or a ready-made product path under /api/v1/*) with Authorization: Bearer <your secret>. The same balance powers HTTP calls and signed-in browser tools, so you can compare UI results against your backend without maintaining two wallets.

Why get a Jev API key from this host

Official TypeSafe access, OpenRouter, Vercel AI Gateway, and Cloudflare AI remain valid paths when you already have upstream credentials or can wait on a product waitlist. Many agent teams need to ship routing, triage, or moderation this week without that access. This host trades a retail markup for instant issuance, prepaid metering, and production-shaped templates. Use official docs for deep primitive theory and brand-level product news; use this page when the blocking question is simply how to obtain a callable credential and hit decide. If you later receive an official key, you can migrate model calls upstream and keep this host only for templates you still find useful — nothing here claims to replace TypeSafe’s commercial product.

Prepaid packs for your Jev API key

Retail metering is about $0.20–$0.42 per million input tokens depending on pack size. Start with a five- or twenty-dollar pack to validate schemas and latency, then move to a larger Production pack when traffic is real. Output remains free on this host, so asking several questions in one decide call stays cheap relative to chat-model round trips that bill both directions. Packs are prepaid capacity for this host’s decide surface and tools — not transferable TypeSafe credits and not a guarantee of identical upstream latency during model or gateway changes.

• $5

  ≈ 11.9M input tokens

  $0.42/M

• $20

  ≈ 47.6M input tokens

  $0.42/M

• $100

  ≈ 400.0M input tokens

  $0.25/M

• $500

  ≈ 2.50B input tokens

  $0.20/M

What a Jev API key (jv_live_) authorizes

A jv_live_ secret authenticates you to this host’s decide surface. It does not grant downloadable weights, local inference, or a pass-through TypeSafe SLA. Requests accept a state (string or structured JSON) plus questions that use Choice, Score, or Noul primitives. Responses return typed answers with calibrated scores you can branch on in code — for example route billing tickets one way and escalate low-confidence cases another. Keep one active secret per environment (local, staging, production) and rotate if it leaks. If a secret is compromised, revoke it on Account and create a replacement before continuing traffic.

Common Jev API key pitfalls

• Putting the secret in frontend code or public repos — treat it like a Stripe secret key and load it only on the server.
• Expecting chat-style completions — decide returns structured answers, not essays, so prompt as questions with criteria rather than open-ended chat.
• Burning the monthly free grant on load tests — use prepaid packs for anything beyond smoke, and cap concurrency in your client.
• Ignoring confidence — low-confidence answers should fall back to a human queue or a slower generative model instead of forcing automation.
• Assuming official list price ($0.042/M upstream) equals this host’s retail packs — you pay for instant keys and hosted product paths, not a wholesale passthrough.
• Sending huge unstructured blobs as state — trim to the fields your questions need so you do not waste input capacity on noise.

Tools vs calling decide with your Jev API key

Browser tools are the fastest way to validate a question schema on real traffic samples. Once the answers look right, copy the same questions into POST /api/v1/decide from your backend. Batch decide helps when you have many independent states — for example labeling a CSV of support tickets overnight. The free /api/v1/gate helper is useful when you only need a confidence threshold without a full pack of questions. Visual editor, batch UI, and templates exist so you can iterate without rebuilding forms each time. A practical pattern is: prototype in Tools → lock the questions → call decide from CI or your agent runtime with the same JSON.

Keep your Jev API key safe in production

Store secrets in your platform’s secret manager. Prefer short-lived server-side calls over exposing the key to browsers or mobile apps. Cap retries and circuit-break on 429/5xx so a buggy loop cannot drain prepaid balance overnight. Review Account usage if latency or volume spikes unexpectedly. Log request IDs without logging the Bearer token. If you already hold an official upstream key, prefer that path for pure model access; keep this host when you specifically need instant hosted decide plus templates. Rotate secrets after staff changes the same way you would rotate any other production credential.

First decide call with a Jev API key

Replace the placeholder with your secret. The samples below hit the same endpoint your tools use after sign-in. Start with a tiny state and one Choice question; expand questions only after the happy path returns in a few hundred milliseconds.

curl https://jevtypesafe.org/api/v1/decide \
  -H "Authorization: Bearer jv_live_YOUR_KEY_HERE" \
  -H "Content-Type: application/json" \
  -d '{"state":"Customer: my card was charged twice",
       "questions":{"topic":{"type":"choice",
         "instructions":"Route this message",
         "criteria":{"billing":"money","bug":"broken","account":"login"}}}}'

import os, requests

r = requests.post(
    "https://jevtypesafe.org/api/v1/decide",
    headers={"Authorization": f"Bearer {os.environ['JEV_API_KEY']}"},  # jv_live_...
    json={
        "state": "Customer: my card was charged twice",
        "questions": {
            "topic": {
                "type": "choice",
                "instructions": "Route this message",
                "criteria": {"billing": "money", "bug": "broken", "account": "login"},
            }
        },
    },
)
print(r.json())

const r = await fetch("https://jevtypesafe.org/api/v1/decide", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.JEV_API_KEY}`, // jv_live_...
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    state: "Customer: my card was charged twice",
    questions: {
      topic: {
        type: "choice",
        instructions: "Route this message",
        criteria: { billing: "money", bug: "broken", account: "login" },
      },
    },
  }),
});
console.log(await r.json());

Try tools after you have a Jev API key

Run a production template in the browser, then ship the same questions over decide. Support triage and content moderation are the usual first stops because their criteria maps are easy to adapt.

• Support Triage

  Route a support ticket — team, issue type, severity, urgency, and escalate.

• Content Moderate

  Moderate user text — allow, review, or block with category flags.

• Lead Qualify

  Qualify an inbound lead — ICP fit, segment, buying intent, routing.

• Email Triage

  Classify an inbound email — category, priority, spam, needs-reply, routing.

Jev API key FAQ

Where can I get a Jev API key?

On this site: sign in, buy a prepaid pack, create a jv_live_ secret on Account, then call /api/v1/decide. Official TypeSafe product access remains a separate path via typesafe.ai.

Is there a free tier?

Yes for evaluation: anonymous tool runs plus a small monthly grant after registration. Production traffic needs prepaid packs.

Do tokens expire?

Prepaid input capacity stays on your account while the account remains active. Consumed capacity is not refunded.

Can I download model weights and run locally?

No. Jev is API-only on known public channels. This host likewise provides remote decide access, not weights.

Why is retail higher than TypeSafe’s published $0.042/M?

You pay for instant issuance, prepaid metering, hosted decide/batch/gate, and templates — not a pass-through of upstream list price.