Skip the install. Get this working in under 2 minutes.
Start a free trial on cloud.anythingmcp.com, add the Jev by TypeSafe in one click, then point your AI client (Claude, ChatGPT, Copilot or Cursor) at the generated MCP endpoint. No Docker, no git clone, zero engineering experience required.
Summary
A tested recipe: Claude asks TypeSafe Jev three typed questions per ticket and gets urgency, owning team and frustration with probabilities in 0.3 s.
Try asking
Example prompts for Jev by TypeSafe
Click any prompt to copy it. Paste into Claude, ChatGPT, Cursor, Gemini, Copilot or OpenClaw to run it against this connector.
Claude is AI and can make mistakes. Please double-check responses.
Ticket triage with Jev
Short answer: connect Jev to Claude through AnythingMCP, paste your tickets (or let Claude fetch them from your helpdesk connector) and ask Claude to triage them with Jev. Claude sends each ticket to Jev with three questions, urgent or not, which team, how frustrated, and gets back probabilities it can sort and filter on. Clear cases are routed; unsure ones are left for a person.
Triage is the textbook job for Jev: the answer always comes from a fixed list, the same questions apply to every ticket, and you want to know when the model is unsure. A chat model can do it too, but it answers in prose, costs more per ticket and does not tell you how sure it is.
The three questions
| Question | Type | Options or levels |
|---|---|---|
| Does the ticket report a problem that blocks the customer's business right now and needs a same-day response? | yes/no | probability of yes |
| Which team should handle it? | choice | billing (payments, payouts, refunds, invoices), technical (bugs, crashes, outages, integrations), sales (pricing, plans, discounts, new accounts), none (no action needed or nothing above fits) |
| How frustrated is the customer? | score | not frustrated, mildly frustrated, frustrated, very frustrated, each with a one-line description |
The none option matters: without it, a thank-you note would be forced into one of the teams.
A real run
We gave Claude these four tickets and the prompt below. Claude read jev_playbook, fixed its thresholds (unsure means a yes/no between 0.15 and 0.85, or a confidence of 0.6 or lower), then made one jev_ask call per ticket.
Use the Jev tools to triage these tickets: urgency (yes/no probability), owning team (billing, technical, sales, none) and frustration on four levels. Give me a table and flag every answer where Jev was unsure.
| Ticket | Urgent (P yes) | Team (confidence) | Frustration (0 to 3) |
|---|---|---|---|
| "My payouts have been failing for 3 days and my suppliers are waiting." | 0.88 | billing (1.00) | 3 |
| "Do you offer a discount for annual plans? Just curious." | 0.02 | sales (1.00) | 0 |
| "Die App stürzt bei jedem PDF-Upload ab, seit gestern Abend. Wir können nicht arbeiten!" | 0.90 | technical (1.00) | 3 |
| "Thanks for the great webinar last week :)" | 0.01 | none (1.00) | 0 |
Nothing fell into the unsure band. Claude pointed out that the payout ticket, at 0.88, was the closest to it, and suggested handling the German crash report and the payout issue first.
What the call looks like
This is the jev_ask call Claude wrote for the first ticket, unedited:
{
"state": {
"ticket": {
"text": "Help! My payouts have been failing for 3 days and my suppliers are waiting."
}
},
"questions": {
"urgent": {
"type": "noul",
"instructions": "Does `ticket.text` report a problem that blocks the customer's business right now and needs a same-day response?"
},
"team": {
"type": "choice",
"instructions": "Which team should handle `ticket.text`?",
"criteria": {
"billing": "payments, payouts, refunds, invoices",
"technical": "bugs, crashes, outages, integrations",
"sales": "pricing, plans, discounts, new accounts",
"none": "no action needed or nothing above fits, e.g. thanks or praise"
}
},
"frustration": {
"type": "score",
"instructions": "How frustrated is the customer who wrote `ticket.text`?",
"criteria": [
"not frustrated: neutral, curious or happy",
"mildly frustrated: minor annoyance",
"frustrated: clear complaint, problem affecting them",
"very frustrated: distressed, urgent pleas, business blocked"
]
}
}
}
All three questions run in parallel against the same ticket, and the ticket is billed once.
Make it yours
- Fetch instead of paste. Install your helpdesk connector (Zendesk, Freshdesk, Help Scout and others are in the catalog) on the same AnythingMCP server and ask Claude to "triage today's open tickets with Jev".
- Use your own teams. Replace the options with your queues and give each a one-line description of what belongs there.
- Add your rules as evidence. Put the customer's plan or SLA next to the ticket, for example
{"ticket": ..., "customer": {"plan": "enterprise"}}, and refer to it in the question. - Keep people on the unsure ones. Ask Claude to route only answers above your threshold and list the rest for review.
FAQ
How much does triaging a ticket cost? Each ticket above was one Jev call of about 500 input tokens with three questions, which at TypeSafe's $0.042 per million is about two thousandths of a cent. Output is free.
Does it work with tickets in other languages? Our German ticket was routed correctly with full confidence. English is where Jev is most accurate, so check confidence values for other languages.
Can this run without a chat? The recipe is written for Claude, ChatGPT or another MCP client. Every call goes through AnythingMCP and is logged, so the same tool can also be called by any agent that speaks MCP.
Related
- Jev MCP server: the overview
- Connect Jev to Claude
- Connect Jev to ChatGPT
- Jev by TypeSafe in the marketplace
Was this guide helpful?
