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

jev-code

Gives Claude Code, Codex, Pi and OpenCode five typed Jev tools (classify, check, score, rank, ask) plus a skill on when to use them.

MCP serverbeta
View on GitHub ↗

Overview

jev-code turns Jev into a first-class tool for four coding agents. The motivation is that coding agents make many small classification decisions, such as which of 40 CI failures are real, which files answer a question, whether a PR does what its description says, and how severe each finding is, usually by eyeballing, writing regexes or spending a frontier-model call per item.

It provides five tools with the same names and JSON in every harness: jev_classify, jev_check, jev_score, jev_rank and jev_ask. Claude Code, Codex and OpenCode connect over MCP stdio, Pi through a native extension, and any shell through the jev-code CLI. A skill tells the agent when a task needs a classifier, how to write good classes and questions, and how to act on answers, with a reference adapted from TypeSafe's official skill. It supports Jev hosted on TypeSafe, OpenRouter or the Vercel AI Gateway, makes one request per tool call, documents the hosts it talks to in SECURITY.md, and offers one-command setup.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich label applies to this item (e.g. real or flaky CI failure)?—
Yes / noDoes this PR do what its description says?yes · no
ScoreHow severe is this finding?—

How it works

  1. Install jev-code into the agent with one command
  2. Skill guides the agent to call a Jev tool when needed
  3. Tool sends one request to the configured provider
  4. Agent acts on typed answers and probabilities

Strengths

  • Same tools across four agents and the shell
  • Skill teaches good question design
  • Documented network hosts

Caveats

  • Requires a TypeSafe, OpenRouter or Vercel key
  • Agent must choose to use the tools

At a glance

Type
MCP server
Status
beta
Language
TypeScript
Runtime
MCP server, Pi extension, CLI
Has code
yes

Use cases and integrations

How curated lists describe it

  • Jev, TypeSafe.

    wh000wh000/awesome-jev-live

  • Jev, TypeSafe's System One classifier, as a tool inside Claude Code, Codex, Pi, and OpenCode: typed classify, check, score, rank, and ask, plus one-command setup.

    hellogumbo/awesome-jev

  • MCP server giving Claude Code, Codex, Pi and OpenCode five typed tools - classify, check, score, rank, ask - with one request per tool call and the hosts it talks to written down in `SECURITY.md`.

    MrJev/awesome-jev

Frequently asked questions

Which tools are included?

jev_classify, jev_check, jev_score, jev_rank and jev_ask.

Which agents are supported?

Claude Code, Codex, OpenCode (MCP) and Pi (native extension), plus a CLI.

Which providers work?

TypeSafe, OpenRouter or Vercel AI Gateway.

Related projects

  • jev-test-filter

    CLI that scores every test against a git diff with Jev and emits filter arguments for vitest, jest, node, Playwright, cargo, go and rspec.

    ProjectbetaTypeScript
  • jev-tools

    Six evidence-oriented Jev tools for pi, omp and any MCP client: ask about code, find files, locate ranges, check diffs and select tests.

    ServicebetaTypeScript
  • decisionkit

    CLI, MCP server and Claude Code plugin for closed-form yes/no, choice, score and rank decisions via Jev, OpenAI Decisions or Claude.

    MCP serverbetaTypeScript
  • jev-flash-router

    MCP server exposing Jev Noul, Choice and Score via OpenRouter so coding agents can make quick decisions without reasoning tokens.

    MCP serverbetaTypeScript
  • jev-mcp-server

    PyPI MCP server for Jev with choice, score, noul, compare, verify and batch classify tools plus a one-command client installer.

    MCP serverbetaPython
  • hermes-jev-plugin

    Hermes Agent plugin exposing Jev as four coding-workflow tools: jev_check, jev_route, jev_score and jev_evaluate.

    PluginbetaPython