Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

jev-agent-kit

jevkit: zero-dependency CLI and eleven MCP tools giving agents typed Jev decisions for routing, triage, guarding, grep, rank and compaction.

MCP serverbetaMIT
View on GitHub ↗

Overview

jevkit is an unofficial command-line tool and MCP server, published on npm as @walidboulanouar/jevkit, that gives agents fast typed decisions from Jev in one binary with no dependencies on Node 18+. The same tools work as MCP tools inside an agent and as shell commands in pipelines: check (yes/no), choose, score, judge (several yes/no checks that all must pass, a cheap done-check), route (pick the best model, skill or tool, abstaining when nothing fits), triage (label many items), guard (allow, ask or deny an action), and others for grep, rank and compact.

The author notes it pays off when decisions run in batches or gates, such as many lines or a check on every tool call, rather than for single one-off questions an LLM agent can answer itself. A doctor command verifies the key, JEV_MODEL pins the model (recommending jev-1.13.0 for stable behaviour), and JEV_API_URL overrides the endpoint but must be HTTPS except on localhost. It is MIT licensed.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich model, skill or tool best fits this task (or none)?—
ChoiceShould this action be allowed, asked about or denied?allow · ask · deny
Yes / noDo all done-checks pass for this output?yes · no

How it works

  1. Install via npx or npm and set TYPESAFE_API_KEY
  2. Run jev doctor
  3. Call tools from the shell or via MCP
  4. Branch on probabilities, labels or allow/ask/deny

Strengths

  • No dependencies; one binary
  • Same tools in shell and MCP
  • Route abstains when nothing fits

Caveats

  • Unofficial, not affiliated with TypeSafe
  • Early version (0.2.0)

Getting started

`export TYPESAFE_API_KEY=...` then `npx @walidboulanouar/jevkit@0.2.0 doctor`.

At a glance

Type
MCP server
Status
beta
Language
JavaScript
Runtime
Node CLI, MCP server
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • jevkit: fast typed decisions for agents. CLI and MCP tools (route, triage, guard, grep, rank, compact, judge) on TypeSafe Jev. Zero dependencies.

    logicrw/awesome-jev-projects

Frequently asked questions

What tools does jevkit provide?

Eleven, including check, choose, score, judge, route, triage, guard, grep, rank and compact.

When is it worth using?

For batches and gates that run often; not for single one-off questions.

Should I pin the model?

Yes, the README suggests pinning jev-1.13.0 for stable behaviour.

Related projects

  • master-jev-hook

    Local gateway, MCP server and Claude Code hooks that let coding agents delegate explicit choices, classification and scoring to Jev.

    MCP serverbetaTypeScript
  • jev-decision-layer

    MIT plugin and local MCP server that hands an agent's closed questions to Jev or local Laya, with policy gates and receipts.

    Agent skillproductionPython
  • jev-awesome-skills

    Open-source Jev skills for Claude Code, Codex, Cursor and Grok: one client to gate, route, guard, verify, triage, review and act.

    ProjectbetaTypeScript
  • `kaijia323/dsh-plugin-jev`：DeepSeek Harness 的 Jev 插件（与本库同环境）

    DeepSeek Harness plugin that registers a native jev_decide tool and declares its own policy for when the agent should ask Jev.

    PluginbetaJavaScript
  • Jev DSH Decision Engine

    Decision plugin for agent harnesses where Jev recommends tools, skills or owners and scores output, for DeepSeek Harness and iPolloWork.

    Agent skillbetaTypeScript
  • jev-codex-plugin

    Codex plugin that consults Jev to choose tools, models and skills, diagnose failed commands and check completion claims.

    PluginbetaTypeScript