Jev Pricing: Pay Only for Input Tokens

Jev pricing on Jev Agent is simple: you will pay only for the input tokens Jev reads, and every output token will be free, because Jev returns a typed decision instead of a block of generated text. Below you can compare monthly plans, yearly plans and token packs side by side, check the effective cost per 1k decisions for each tier, and run your own workload through an LLM cost comparison calculator before you decide what to buy. This page is for developers evaluating a Jev AI agent or a direct integration with the Jev API who want to know the real cost of typed decisions before writing any code. Early access has not opened yet, so clicking Checkout does not charge you today — it adds you to the early access queue, and Jev Agent will email you as soon as your spot opens.

Free Developer Tier

Every account will get a free developer tier once early access opens: 1.5M input tokens a month, no credit card required. That will amount to roughly 5,000 typical decisions at 300 tokens each — enough to wire up a support router, a ticket triage flow, or a first Jev AI agent and see real usage before you commit to a paid plan. Once it's live, the free tier will renew automatically every month, will never require a subscription, and will use the same request format as every paid plan, so nothing in your integration will need to change when you outgrow it and move to a monthly or yearly plan. There will be no trial clock counting down — the free developer tier will stay free for as long as your monthly usage fits inside 1.5M input tokens.

Start Free Developer Tier

Switch between the Monthly, Yearly and Packs tabs above to see how each billing cadence prices out. Yearly plans cut the annualized monthly rate in half, so they suit teams that already know roughly how much volume they need for the year. Packs are one-time purchases for spiky or seasonal usage, with no subscription attached. The Max plan and the 600M token pack both include a quantity slider so you can scale up to five times the base tier — the total price and included tokens update as you drag it, without switching to a different plan.

Rate Card and Cost per 1k Decisions

The table below lists every plan's price, included tokens, effective rate per million tokens, and cost per 1k decisions at 300 tokens each, so you can compare tiers without doing the math yourself. Larger plans and larger packs bring the per-token rate down, but every tier will bill only the input tokens Jev reads — there is no separate charge for output, and no per-request or per-seat fee layered on top. Use this table to sanity-check the calculator's recommendation below, or to price out a workload you already understand in decisions per month rather than raw tokens.

LLM Cost Comparison Calculator

The calculator's starting numbers — decisions per month, tokens per decision, an LLM's input price and its output token count — are editable assumptions, not measurements of your workload. Adjust them to match how your product actually calls a model: a support-ticket classifier might average 300 input tokens and no output at all, while a longer agent prompt could run much higher. TypeSafe reports that comparable LLM APIs price input at $0.20–$10 per 1M tokens, with output tokens typically costing about five times as much as input (TypeSafe blog), so set the slider to whatever model you would otherwise call. The calculator then recommends the cheapest Jev Agent plan or pack for that volume and estimates the percentage you would save by moving the decision itself off the LLM and onto Jev. Because Jev Agent will only bill for input, workloads that currently lean on long LLM outputs — summaries, drafts, explanations — tend to show the widest gap between the two monthly costs.

What Counts as a Token

Jev Agent will only count the input tokens Jev reads when it evaluates a state against your Choice, Score or Noul questions — output will never count, because Jev does not generate text. A 250-word support ticket sent as state will cost only the tokens in that ticket and your question map, not a single token for the answer that comes back. That is a meaningful difference from typical LLM billing, which usually charges for tokens in both the prompt and the completion, regardless of how short the useful part of the answer is. Once paid plans open, monthly plans will reset their token balance every billing month; yearly plans will grant the full twelve months of tokens upfront in one deposit, valid for 365 days from purchase. Token packs will be a one-time purchase that never expires, so you can buy a pack for a seasonal spike and draw it down whenever it fits, on top of a plan or on its own. None of these are TypeSafe's own prices: they are Jev Agent's rates for reselling access to Jev, and a plan here will provision a Jev Agent key rather than an official TypeSafe API key.

Next Steps

Ready to see the numbers against your own product? Get Started with Jev API to read the request and response shapes before you commit to a plan, or Launch Jev Agent Builder to step through the decision loop an illustrative Jev AI agent runs. If you would rather zoom back out first, Return to Jev Overview for the full picture of what Jev and Jev Agent do, then come back here once you know roughly how many decisions a month you expect to make. Common billing questions — the free tier, LLM cost comparisons, and whether this is TypeSafe's own pricing — are answered just below.