nexibeo

Jev Cookbook — Data skill for Claude Code

Data community

Practical, tested recipes for TypeSafe's Jev decision model on OpenRouter: support triage, database indexing, file organizing, tagging, taxonomies, dedupe, PII detection, extraction, search re-ranking.

How to install Jev Cookbook

This entry records only its repository, not the path inside it, so there is no exact command to give. Open nexibeo/jev-cookbook and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Jev Cookbook does

Alternatives in Data

Clinical Decision Support — Evidence-based clinical decision support tools 6.4k ★
Gbp — Manage the Google Business Profile, live review triage, public review replies, Google posts, analytics 195 ★
Indonesia Gov APIs — 🇮🇩 50+ Indonesian Government APIs & Data Sources — BPS, OJK, BPJPH, BPOM, Bank Indonesia, IDX, BMKG + MCP se 134 ★

README

Jev Cookbook

**Practical, tested recipes for [TypeSafe's Jev](https://docs.typesafe.ai/introduction), the fast decision model on [OpenRouter](https://openrouter.ai/typesafe/jev-1.13).** Fifteen real-world jobs, each with a runnable script, a small labelled dataset and measured results: support triage, database indexing, a file organizer, tagging, category trees, duplicate detection, PII scanning, bank transactions, invoice extraction, search, log triage, moderation, lead scoring, a browser agent and a Gmail labeler that connects to your own inbox.

**Want Jev in a real browser?** [Jev Browser Control](https://github.com/nexibeo/jev-browser-control) is the sibling project: an MCP server that lets Claude or Codex drive Chrome, with Jev picking every click, keystroke and scroll. Recipe 14 below is the bare loop; that repo is the finished tool.

Created by **[Jeroen Erne](https://www.linkedin.com/in/jeroenerne/)** ([nexibeo.com](https://nexibeo.com) · [completeaitraining.com](https://completeaitraining.com)), built together with Claude.

What Jev is, in 30 seconds

Jev is not a chat model. You send it some data (the `state`) and a set of questions with fixed answers, and it returns typed answers with probabilities, usually in under half a second.

Choice: pick one option from a list you define (up to 255), with a probability for each.
Noul: the probability that a yes/no statement is true.
Score: a position on levels you describe, lowest first.

It never writes text, so it can't hallucinate an answer outside your options, and it is cheap: $0.042 per million input tokens, output free. Across these recipes that's **2 to 8 cents per 1,000 items**. It can still pick the wrong option, which is why every recipe ro