jev (shaharia-lab)

Rust CLI for Jev: ask yes/no, choice and rubric questions about text and branch on exit codes, JSON or MCP tools.

MCP serverbeta

Overview

jev-cli by shaharia-lab is a community-built command-line tool for TypeSafe's Jev model, described as a semantic if statement for the terminal. You send some content and a question you define, get calibrated probabilities back, and branch on them in a shell script, a CI job or an AI agent. Answers come out as exit codes for shells and CI, JSON for scripts, and MCP tools for agents.

A distinctive feature is the exit code design: it separates a false answer from an answer that falls inside your abstain band, so a script can take a third branch and ask a person. It also lints the request before spending anything. The README's examples include triaging support tickets, routing incoming messages and moderating user content with borderline cases sent to people. It is published on crates.io with GitHub releases, and the README states clearly that it is unofficial and not affiliated with TypeSafe. A TypeSafe API key is required, and the README's cost figures are per-request estimates.

What the decision model decides

Type	Question	Answers
Yes / no	User-defined yes/no question about the input text	yes · no
Choice	User-defined multiple-choice question	—
Score	User-defined rubric question	—

How it works

Define a question and its answers
Pass text with --state or from stdin
Jev returns calibrated probabilities
Branch on exit code, JSON or MCP tool output
Route abstain-band answers to a person

Strengths

Exit codes distinguish no from unsure
Request linting before spending
Works in shells, CI and agents via MCP

Caveats

Unofficial community project
Requires a TypeSafe API key
Thresholds and abstain bands are yours to tune

Getting started

Install the jev-cli crate or a release binary, set your TypeSafe key, and run jev noul "Is this customer angry?" --state "..."

At a glance

Type: MCP server
Status: beta
Language: Rust
Runtime: CLI and MCP server
Has code: yes
Has tests: yes

Use cases and integrations

Infra / SDKs / Integrations537

How curated lists describe it

Command-line tool for TypeSafe AI's Jev model. Ask yes/no, multiple-choice and rubric questions about any text and get calibrated probabilities back. Answers become exit codes for shells and CI, JSON for scripts, and MCP tools for AI agents.
hellogumbo/awesome-jev
Rust CLI whose exit codes separate a false gate from an answer inside your abstain band, so a script can take a third branch and ask a person. Lints the request before it spends anything.
MrJev/awesome-jev

Frequently asked questions

What does Jev decide?

Whatever typed question you define: yes/no, multiple choice or a rubric score over your text.

How does a script handle uncertain answers?

A distinct exit code marks answers inside your abstain band, so the script can ask a person.

Is it official?

No. The README says it is community-built and not affiliated with TypeSafe.

Related projects

daf-jev
Composable Python toolkit for Jev: question builders, confidence gates, batch evaluator, calibration, CLI, MCP server and agent skill.
MCP serverbetaPython
jev-mcp (freepik-company)
Go MCP server giving any client typed Jev decisions via OpenRouter or TypeSafe, with validated answers and no paid retries.
MCP serverproductionGo
askjev
Unofficial hosted or local Jev MCP server giving agents a quick second opinion: it picks yes/no, scale or choice and returns probabilities.
MCP serverbetaTypeScript
fuzzy-jev
Rust CLI and library that asks Jev typed questions via OpenRouter and combines the probabilities with fuzzy-logic rules into decisions.
CLIbetaRust
jev (drpaneas)
Unofficial standard-library Go client for Jev with generic typed Choice values and local confidence thresholds.
ProjectbetaGo
Jev Studio
pip-installable Jev kit: a CLI for verify, screen, classify, extract, route and rerank, plus an MCP server for the primitives.
MCP serverbetaPython