Jev in n8n
n8n workflows are full of small decisions — which branch does this ticket take, is this lead worth scoring, should this post go live or to review. Jev answers each as a typed value with calibrated confidence in one call, so you branch on a key you can trust instead of parsing an LLM's prose. You don't even need a custom node: the built-in HTTP Request node is enough.
Cada ejemplo necesita una jv_live_ key. Consigue una en la página de precios, expórtala como JEV_API_KEY y ya estás listo. Obtener una key →
Jev is TypeSafe AI's System One decision model: send state plus typed questions, get back one option per question (choice), an ordered rating (score), or a calibrated yes/no (noul), each with a probability. In n8n that maps perfectly onto a Switch — one decision call, one branch per answer.
Fastest path: the built-in HTTP Request node
No install, nothing to maintain. Add an HTTP Request node and POST to the hosted gateway with your jv_live_ key — every decision then bills through your own Jev account:
Method: POST
URL: https://jevtypesafeai.com/api/v1/decide
Auth: Header → Authorization: Bearer {{$env.JEV_API_KEY}} // jv_live_...
Headers: Content-Type: application/json
Body (JSON):
{
"state": "={{ $json.ticket }}",
"questions": {
"queue": { "type": "choice", "instructions": "Which queue should this go to?",
"criteria": { "billing": "payments", "technical": "bugs", "abuse": "spam/fraud", "sales": "pricing" } },
"urgent": { "type": "noul", "instructions": "Does this need a human within the hour?" }
}
}The response is a typed map. Read it in any downstream node or expression:
{{ $json.answers.queue.choice }} // one of your criteria keys
{{ $json.answers.queue.confidence }} // calibrated 0–1 — threshold for a fallback
{{ $json.answers.urgent.noul }} // calibrated yes/no probabilityBranch on it with a Switch node
- Add a Switch node after the HTTP Request node.
- Set its value to {{ $json.answers.queue.choice }} and add one rule per option (billing / technical / abuse / sales).
- Add a fallback route for {{ $json.answers.queue.confidence }} below your threshold — the low-confidence tail goes to a human.
- Everything downstream branches on a key that's guaranteed to be one of yours — never a hallucinated label.
Or use a community Jev node
Several community nodes wrap this with a nicer UI — a 'Classify & Route' operation that creates one output branch per answer plus a 'Needs Review' branch below your confidence threshold, like a typed Switch. Install one, then point its base URL at the hosted gateway and use your jv_live_ key so usage bills through your account:
# In n8n: Settings → Community Nodes → Install
n8n-nodes-jev
# Then set the node's endpoint/key to the hosted gateway:
# Base URL: https://jevtypesafeai.com/api/v1/decide
# API key: jv_live_...Community nodes default to the official api.typesafe.ai endpoint, so set the base URL and a jv_live_ key explicitly if you want decisions to run through your hosted account.
What to automate with it
- Ticket & email triage — route each item to the right queue, send the unsure ones to review.
- Lead scoring — rate inbound leads on an ordered scale and branch the hot ones to sales.
- Content moderation gate — a calibrated noul before a post publishes, with a human branch below threshold.
- Agent steps — let Jev pick the next tool or action inside an n8n AI workflow.
Preguntas frecuentes
Do I need a custom node to use Jev in n8n?
No. The built-in HTTP Request node is enough — POST to https://jevtypesafeai.com/api/v1/decide with your jv_live_ key and read answers.<question>.choice / .noul / .score from the response. Community nodes add a nicer Classify & Route UI, but the HTTP node works everywhere with nothing to install.
How do I branch on a Jev decision in n8n?
Add a Switch node after the decision call and set its value to {{ $json.answers.queue.choice }}, with one rule per option plus a fallback for low {{ $json.answers.queue.confidence }}. Because the choice is locked to your criteria, every branch maps to a valid key and the uncertain tail routes to a human.
Which key do I use, and does it bill through my account?
Use a jv_live_ key from jevtypesafeai.com/pricing and point the call (or a community node's base URL) at https://jevtypesafeai.com/api/v1/decide. Then every decision runs through your hosted account — per input token, output free. Community nodes default to api.typesafe.ai, so set the base URL explicitly.
Are there community n8n nodes for Jev?
Yes — several (for example n8n-nodes-jev) add a Jev node with a Classify & Route operation that creates one branch per answer plus a Needs Review branch below your confidence threshold. Install from Settings → Community Nodes, then set the endpoint and key to the hosted gateway.
What's a good first workflow?
Ticket or email triage: an HTTP Request node asks Jev a choice question for the queue and a noul for urgency, a Switch routes by answers.queue.choice, and anything below your confidence threshold goes to a human review branch. From there, lead scoring and content gating follow the same shape.
Ver también: How to use the Jev API · Jev MCP server · Jev as a judge · Pricing
Prueba Jev antes de integrarlo
Ejecuta una decisión real y tipada en el navegador — gratis, sin registro — y luego pon tu key en el ejemplo de arriba.