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

multi-agent-jev

Rust template for multi-agent orchestration with Jev as the L0 layer for typed routing, risk and escalation decisions.

Projectprototype
View on GitHub ↗

Overview

multi-agent-jev is a reusable template, documented in Portuguese, for multi-agent software workflows that use Jev as a fast L0 decision layer. A jev-core crate handles routing, risk, escalation and briefs as typed decisions, an orchestrator crate drives the flow through git worktrees, and agent contracts define typed artifacts such as DesignDoc and ReviewReport.

The layers are L0 decision, L1 design (Maestro), L2 implementation (Codesmith), L3 adversarial review (Codex) and L4 merge, where a human is the only merge gate. Principles include design before code, AGENTS.md as the canonical contract, Jev producing only typed schemas, and AI never deciding irreversible actions. It is intended for teams who want a structured starting point for agent pipelines.

What the decision model decides

TypeQuestionAnswers
ChoiceRoute this task to which agent or flow(inferred)—
ScoreRisk level of this task(inferred)—
Yes / noEscalate to a human?(inferred)yes · no

How it works

  1. Task enters jev-core for route, risk, escalate and brief decisions.
  2. Orchestrator moves it through design, implementation and adversarial review in worktrees.
  3. Thresholds and critical paths configured in jev-core decide escalation.
  4. A human approves the merge on a protected main branch.

Strengths

  • Clear separation of decision, design, implementation, review and merge.
  • Typed outputs only from Jev.
  • Human remains the sole merge gate.

Caveats

  • Template; thresholds and paths must be configured per project.
  • Docs are in Portuguese; no licence stated.

Getting started

cargo test --workspace --tests && cargo build --workspace, then copy templates/project/ into a new repository.

At a glance

Type
Project
Status
prototype
Language
Rust
Runtime
CLI orchestrator
Has code
yes
Has tests
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

What does Jev do here?

It is the L0 layer that makes typed routing, risk, escalation and brief decisions.

Can AI merge code?

No; the human is the only merge gate and AI never decides irreversible actions.

How do I adopt it?

Copy templates/project/, adapt AGENTS.md, set thresholds in jev-core and enable branch protection.

Related projects

  • openjev

    opencode plugin that routes the agent's small multiple-answer decisions to Jev and labels each answer auto or escalate by confidence.

    PluginbetaTypeScript
  • TriRouter

    Prompt router for Claude Code, Codex and Antigravity that picks model, effort, skill and language per request via Jev.

    ProjectbetaPython
  • jev-research

    Guide and small prototype for using Jev with Herdr to route tasks to coding agents like Claude, Codex and Hermes under code-owned policy.

    Projectprototype
  • LangChain

    LangChain's Python framework, which includes an optional TypeSafeClassifier runnable plus model-routing and risky-tool middleware using Jev.

    PluginproductionPython
  • LangChain.js

    LangChain.js, which includes an optional TypeSafeClassifier runnable and middleware for bounded routing and tool-call checks with Jev.

    PluginproductionTypeScript
  • Learn Jev end to end

    Free 12-notebook course building 13 agent tools with Jev as the fast decider and an LLM as the slow writer, for under $0.20.

    ProjectproductionPython