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

jev-agent-harness

Beginner manual, video script and three small examples for building an agent harness with Jev via langchain-typesafe.

Projectprototype
View on GitHub ↗

This repository is mostly documentation or write-up rather than runnable code.

Overview

This repository is a learning guide by Abdul Qaadir based on LangChain's blog post Building a Harness with Jev. It contains a full beginner manual covering Noul, Choice and Score, a ready-to-film ten-minute YouTube script with production notes, and three short Python examples: a minimal Noul call through langchain-typesafe, a ModelRouter middleware demo, and an AutoMode guardrail that checks bash tool calls.

It explains the harness idea that the LLM thinks and writes while Jev makes the structured decisions in the agent loop, with two patterns: ModelRouterMiddleware, which picks a cheap or powerful model per request, and AutoModeMiddleware, which checks every tool call and blocks risky ones before they run. The headline speed and cost multiples it quotes come from the source blog rather than its own measurements. It is mostly documentation.

What the decision model decides

TypeQuestionAnswers
Yes / noDoes this need attention right now?yes · no
ChoiceShould this request go to a cheap or a powerful model?(inferred)cheap · powerful
Yes / noIs this tool call risky?(inferred)yes · no

How it works

  1. Install langchain-typesafe and set TYPESAFE_API_KEY
  2. Run the Noul example
  3. Add ModelRouter middleware for model choice
  4. Add AutoMode middleware to gate tool calls

Strengths

  • Beginner-friendly explanation of harness patterns
  • Runnable minimal examples

Caveats

  • Mostly prose and a video script
  • Performance claims are quoted, not measured here

Getting started

`pip install langchain-typesafe`, `export TYPESAFE_API_KEY=...`, then run examples/noul_classification.py.

At a glance

Type
Project
Status
prototype
Language
Python
Runtime
LangChain
Has code
yes
Has tests
no

Use cases and integrations

How curated lists describe it

Frequently asked questions

What is in this repo?

A manual, a video script and three small langchain-typesafe examples.

What are the two harness patterns?

Model routing middleware and AutoMode tool-call guardrail middleware.

Is it based on another source?

Yes, LangChain's Building a Harness with Jev blog post.

Related projects

  • jev-handson

    Korean course code for attaching Jev to Claude Code, Codex and OpenClaw: skills, hooks, routers and open-model comparisons.

    ProjectproductionJavaScript
  • jev-intro-to-decision-models

    Companion code for the ZazenCodes course on decision models: primitives notebook, use-case examples and a Pi model-router build.

    ProjectproductionPython
  • 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
  • jev-engineering-zh

    Chinese translation of 'Jev Engineering for coding agents', design notes on a harness where Jev makes every per-turn decision.

    Projectdiscussion
  • He says he co-invented ChatGPT. His new AI will not write a word

    Dev.to article on Jev's founder and a walkthrough of using Jev through the Vercel AI SDK's evaluate integration.

    ArticlediscussionTypeScript
  • Jev vs GPT-6 Astra: when to use each

    Vercel's decision guide on when to use a System One model like Jev versus a frontier model like GPT-6 Astra.

    Articlediscussion