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

master-jev-hook

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

MCP serverbeta
View on GitHub ↗

Overview

Master-JEV Hook connects Claude Code, Claude Desktop and Codex to Jev through a local MCP server. Instead of the agent deciding alone between explicit alternatives (which approach, which source, which next action, how to classify or score a batch), it delegates the choice to Jev through MCP tools, and a local HTTP gateway logs every query in a dashboard. Batch tools cover classify (Choice), verify (Noul), score and rank (Score).

For Claude Code only, a set of hooks injects an orchestration rule into the session, announces when Jev is queried, and performs spot checks on risky Bash commands, completion without verification and what to preserve during compaction, always without blocking the session. The gateway uses TypeSafe's official SDK for types but calls the systemone endpoint directly, with OpenRouter or Vercel AI Gateway as alternatives and retries only on rate-limit or overload errors. The README is clear that Jev chooses between options the agent already raised; it does not generate text, prove facts or grant permissions. Documentation is in English and Brazilian Portuguese.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich approach, source or next action should the agent take among these options?—
Yes / noIs this Bash command risky? / Is the task claimed complete without verification?—
ScoreHow should these items be scored or ranked?—

How it works

  1. Agent raises explicit alternatives
  2. MCP tool sends them to the local gateway
  3. Gateway asks Jev and logs the query
  4. Agent acts on the typed answer
  5. Claude Code hooks run non-blocking spot checks

Strengths

  • Works across Claude Code, Claude Desktop and Codex
  • Local dashboard of all queries
  • Non-blocking hooks
  • Multiple provider options

Caveats

  • Agent must choose to call the tools
  • Hooks advise rather than block
  • Requires a provider key

At a glance

Type
MCP server
Status
beta
Language
TypeScript
Runtime
local HTTP gateway, MCP server and Python hooks
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

Does Jev take actions?

No, it chooses among options the agent raised; it does not act or grant permissions.

Do the hooks block the session?

No, they announce, advise and spot-check without blocking.

Which clients are supported?

Claude Code, Claude Desktop and Codex.

Related projects

  • 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-agent-kit

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

    MCP serverbetaJavaScript
  • 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
  • jev-mcp

    MCP server giving any agent Jev's classify, score, yes/no check and allow/confirm/block gating of risky tool calls.

    MCP serverbetaJavaScript
  • 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
  • jevwire

    Jev decision layer for agents: seven MCP tools, an embeddable library and escalate-only Claude Code hooks at harness boundaries.

    MCP serverbetaTypeScript