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

codex-jev-integration

Standard-library MCP server giving Codex advisory Jev tools for role and resource selection, with local validation and 85 offline tests.

MCP serverbeta
View on GitHub ↗

Overview

Codex Jev integration (Jev 2.2) is a standard-library-only MCP server for advisory role and resource selection in Codex CLI and App. It exposes six tools: jev_status, jev_route, jev_report_outcome, jev_select_resources, jev_decide and jev_report_selection_outcome. The repository versions four runtime modules, a configuration example, interface contracts and 85 offline regression tests.

jev_decide shares the task goal, success criteria, environment, constraints, evidence, unknowns and full candidate descriptions across separate Choice questions; missing critical context goes back to Codex. Required resources stay authoritative, and freshness, dependency and conflict checks, response validation and confidence fallback run locally. The tools never execute resources, launch agents, switch models, change permissions or approve actions. Requests go to the fixed TypeSafe endpoint with jev-latest, single-attempt, without redirects and with bounded sizes, sending only sanitized input.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich role or resource best fits this task?eligible candidates

How it works

  1. Codex calls jev_decide or jev_route
  2. Local gates check context and candidates
  3. Jev answers Choice questions
  4. Local validation and confidence fallback
  5. Outcomes reported back for records

Strengths

  • Advisory only; never acts
  • Strict request bounds
  • Offline regression tests

Caveats

  • Machine-specific setup kept outside the repo
  • Caller-reported outcomes are not independent verification

Getting started

python -B payload/jev/decision.py serve --config payload/jev/config.json, registered under [mcp_servers.jev_decision] in Codex config.

At a glance

Type
MCP server
Status
beta
Language
Python
Runtime
MCP server (stdio)
Has code
yes
Has tests
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

Can it run tools or approve actions?

No, it is advisory only.

What is sent to TypeSafe?

Only sanitized input explicitly allowed for transmission.

How is it tested?

85 offline regression tests.

Related projects

  • 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
  • jev-layer

    Portable decision layer for agent harnesses: Jev picks from host-supplied capabilities while the host keeps permissions and execution.

    ProjectbetaTypeScript
  • 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
  • jev-skill-suggester

    Codex skill and Python CLI that recommends one installed skill per task using Jev Choice ranking and Noul fit checks, or returns none.

    Agent skillbetaPython
  • 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