Jev is an AI decision model and API that takes application state plus typed questions and returns structured, code-usable answers with probabilities. Built by Jev AI, it runs on the web with a free online playground and a REST endpoint at POST /v1/systemone, returning results in 70–500ms.
What is Jev?
Jev is a decision tool for software teams that handles classification, routing, scoring, and safety checks. You give it state (a ticket, message, form fields, or structured agent context) and up to eight typed questions, and it returns a typed decision — a choice, score, or yes/no answer — with probabilities included. It is accessed through a free online playground, a REST API, and a Python SDK (typesafe_sdk), and is made by Jev AI.
Key Features
- Typed output shape — Define possible answers up front and receive results as choices, scores, or yes/no values that pass directly into functions, workflows, and databases.
- Multiple questions per request — Ask up to eight typed questions against one shared state, so classification, scoring, and yes/no checks happen in a single call instead of chained requests.
- Probabilities and confidence — Each answer includes probability or confidence values (e.g., 0.94 intent, 0.91 confidence) so code can decide whether to continue automatically or ask a person.
- Fast responses — Decisions return in 70–500ms, suitable for routing, guardrails, and queue logic running many times per second.
- Free online playground — Enter a real scenario, define typed questions, and inspect typed decisions with unlimited free runs.
- REST and SDK integration — A consistent request shape, explicit schemas, and manageable API keys via POST /v1/systemone and a Python SDK using Choice and Noul question types.
- Agent Skill — Install the public Jev Agent Skill into Codex, Claude Code, Cursor, or compatible agents with one API key, supporting English and Simplified Chinese.
Who is it for?
- Backend and platform developers — Route requests to faster or stronger models based on assessed task difficulty, escalating only when needed.
- Agent builders — Add a typed decision layer for routing, guardrails, verification, and completion checks inside coding agents.
- Support and operations teams — Classify intent, urgency, and human-review needs together so tickets take the right path.
- Product engineers handling sensitive actions — Check intent before a tool executes and request confirmation around deletion, payment, or other risky operations.
What can you do with Jev?
- Model routing: Assess task difficulty first, send simple requests to a faster model, and escalate only when needed.
- Tool-call risk gate: Check intent before a tool executes and ask for confirmation around deletion, payment, or other sensitive actions.
- Support triage: Classify intent, urgency, and human-review needs together so tickets route correctly.
- Context compaction: Decide which tool results still matter and preserve source details when long-session context is compressed.
How does Jev work?
- Give Jev the state — a ticket, message, form fields, or the structured context your agent already sees.
- Ask up to eight typed questions describing the decision as a choice, score, or yes/no.
- Let your code act on the result and its probability to queue, route, block, or request review, keeping your business logic in control.
Pricing
Jev's online playground is free with unlimited runs. API pricing tiers are not specified on the page.
