jev-cli
jevctl: Node CLI for typed Jev judgments (verify claims, screen agent hijacks, classify) with exit codes for CI gating.
Overview
jev-cli by Nasrallah-AL, published on npm as jevctl with the command jev, wraps Jev into commands you can pipe into, script around and gate CI on. Commands include verify (does evidence support, contradict or ignore each claim), screen (is this text trying to hijack an AI agent, and is it worth reading), and classify (which label fits, which labels apply, where in a hierarchy), each documented on its own page, among others.
In the first-run example, verifying the claim helmets are optional for adults against evidence that every rider must wear a helmet returns contradicted at confidence 1.00 with an auto action, and the command exits with code 2. Any command supports JSON output and a dry-run that shows exactly what would be sent. The key is entered at a hidden prompt and stored in the OS keychain. It needs Node 20.12+, has CI and a website, and is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Does this evidence support, contradict or not address each claim? | supported · contradicted · unrelated |
| Yes / no | Is this text trying to hijack an AI agent? | yes · no |
| Choice | Which label fits? | — |
How it works
- Install jevctl and log in (key stored in keychain)
- Run verify, screen or classify
- Read verdict, confidence and action
- Use exit codes to gate CI
Strengths
- Exit codes suited to CI gating
- Dry-run shows the exact request
- Keychain key storage
Caveats
- Requires a TypeSafe key
- Verdicts are model judgments
Getting started
`npm install -g jevctl`, `jev auth login`, then `jev verify "<claim>" --evidence "<text>"`.At a glance
- Type
- CLI
- Status
- beta
- Language
- TypeScript
- Runtime
- Node CLI
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
The jevctl command-line tool connects verification, classification and scoring questions to text input and scripts.
logicrw/awesome-jev-projectsCommand-line tool for TypeSafe's Jev AI model.
hellogumbo/awesome-jevTyped judgments from the command line, published to npm as `jevctl`.
MrJev/awesome-jev
Frequently asked questions
What does jev verify do?
Judges whether evidence supports, contradicts or ignores each claim, with exit codes.
What does jev screen check?
Whether text tries to hijack an AI agent and whether it is worth reading.
Where is my key stored?
In the OS keychain.
Related projects
Precision-first RAG service where Jev routes queries, triages each retrieved chunk and verifies every claim before an answer is released.
ServicebetaPythonAdds Jev semantic rules to Zod 4 schemas: Zod checks the shape, Jev checks the meaning, all rules in one request.
LibrarybetaTypeScriptPilot that puts Jev into LangGraph routing and Deep Agents guardrails and claim verification, with executed notebooks and caveats.
ProjectprototypePythonUnofficial Go client for Jev with YesNo, OneOf and Levels constructors and typed answer accessors.
LibrarybetaGoNode MCP server with a dozen fail-closed Jev judgment tools: verify, screen, find, rerank, classify, extract, review and gate.
MCP serverbetaTypeScriptNative macOS agent app wiring 23 LLM providers into a task loop, listed with an optional Jev command-risk advisor via a TypeSafeKit client.
LibraryproductionSwift