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

jev-in-mcp

MCP relay that adds use_jev to every server: Jev picks tool calls, the calling model writes free text, the relay executes.

MCP serverbetaMIT
View on GitHub ↗

Overview

jev-in-mcp places Jev inside an MCP relay. To the agent it is one MCP server; to the user's MCP servers it is a client. All tools pass through unchanged, each server gains an extra __use_jev tool, and a global use_jev covers all servers. Each hands a goal to Jev, which picks the tool calls in a loop (choosing the server first for the global tool), while the calling model supplies any text that must be written and the relay executes.

Setup uses a local page to store the Jev key (OpenRouter or TypeSafe) so it never passes through chat, an import command that copies mcpServers entries from Claude Code, Cursor, Windsurf and Claude Desktop, and a status command. Config supports per-server purposes, denied tools and improved tool descriptions, plus step and result-size limits. It is built on jev-dev-kit and installed from GitHub since it is not yet on npm.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich server and tool should be called next for this goal?—

How it works

  1. Replace client MCP config with the relay
  2. Agent calls use_jev or a server's __use_jev with a goal
  3. Jev picks tool calls; the calling model writes needed text
  4. Relay executes within step and deny limits

Strengths

  • One entry wraps all existing MCP servers
  • Per-tool deny lists
  • Key never passes through chat

Caveats

  • Not yet on npm
  • Jev cannot write values; the calling model must

Getting started

`npx jev-in-mcp setup`, `npx jev-in-mcp import`, `npx jev-in-mcp status`, then use a single jev entry in your MCP config.

At a glance

Type
MCP server
Status
beta
Language
TypeScript
Runtime
MCP relay (Node 20+)
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • MCP relay that adds use\_jev to every server: Jev picks the tool calls, the calling model writes the values Jev cannot choose, the relay executes. Built on jev-dev-kit.

    logicrw/awesome-jev-projects

Frequently asked questions

What does Jev decide?

Which server and tool to call next toward a goal.

Who writes text arguments?

The calling model, since Jev cannot generate text.

Can I block dangerous tools?

Yes, with per-server deny lists.

Related projects

  • 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-skill-router

    Local Python MCP service that keeps the skill catalog out of the agent's context and lets Jev pick which SKILL.md files to load.

    MCP serverbetaPython
  • 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-mcp (burnigtm)

    Local MCP server putting Jev in the coding loop: route next steps, pick prepared tool calls, review patches and verify claims.

    MCP serverbetaTypeScript
  • toolJev

    Code Mode for MCP where retrieval shortlists tools and Jev picks one or says none fits, benchmarked on several tool-use suites.

    BenchmarkresearchPython
  • `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