Will adding a Jev call actually save money or just add latency?

It can, but only if the gate actually skips expensive work. The repo tells you to start in shadow mode, log advice versus normal behavior, then switch to active mode, because active mode saves nothing unless the agent honors the action.

What happens when Jev answers with low confidence?

No. The policy is act at 0.80 or above, surface between 0.50 and 0.79, and escalate below 0.50. Low-confidence answers route to asking a human or running the normal path, and the kill switch always wins.

Can Jev approve an irreversible action on its own?

No. The honest-limits section says Jev never sends, publishes, pays, deletes, or changes permissions, and those stay behind human confirmation no matter what Jev says.

Are the flight example numbers a benchmark I can quote?

Treat them as a method template from one local flight-dates run, not a benchmark or savings guarantee. The repo ships an A/B template so you can measure your own before and after behavior.

What are the two tracks for using it?

Track A is developer: install the TypeSafe SDK, set TYPESAFE_API_KEY and run the router in src/ or the recipes. Track B is the Muse/Hatch agent, where the credential is already connected, the agent calls Jev through its tooling and follows the drop-in skill with no key handling.

What does the core decision loop do?

The agent checks the kill switch, builds a compact state (goal, kind, cache hints, error history, constraints), makes one Jev call with choice, score and noul questions in parallel, then applies policy: act at 0.80 or above, surface from 0.50 to 0.79, escalate below 0.50. The chosen action is logged for calibration.

What recipes ship with it?

Six copy-paste question packs: triage, rank options, act-or-wait, retry gate, research cap and approval gate. The worked example caps flight research at five sources when the noul answers yes at 0.88 and confidence clears 0.80.

CommunityJev Agents & MCP

Muse Jev Playbook

Agent-oriented playbook for using Jev as a cheap, fast decision layer that triages, classifies, scores, and gates work before expensive browser, research, retry, or subagent steps.

See highlights

Open github.com

Stars: 25
Language: Python
License: MIT
Published by: Community
Author: Bodila51
Last verified: Oct 2026

Bodila51/muse-jev-playbook

User repository on GitHub

View on GitHub

Jev decision layer for Muse: a fast, cheap TypeSafe AI gate before expensive agent work — confidence policy, recipes, reference router, honest measurement.

Forks: 3
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Use TypeSafe AI's Jev as a cheap, fast decision layer inside an AI agent workflow — triage, classify, score, and gate work before expensive steps (browser, deep research, retries, subagents). This is a practical, agent-oriented playbook: concepts, a confidence policy, copy-paste question recipes, a drop-in skill, and a real-world case study. It is adapted from grok-bot-jev and generalized for any agent runtime (Muse/Hatch, custom harnesses, or plain scripts).

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import

Works forJev for Classification Jev for Routing Jev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

Three typed primitives, choice, score, and noul, answered in one parallel pass with probabilities and 0-1 confidence.
Two tracks share one policy: developer runtime with SDK and TYPESAFE_API_KEY, or Muse and Hatch via a drop-in skill.
Confidence policy with act, surface, and escalate bands plus shadow logging before active enforcement and a kill switch.
Six copy-paste recipe packs covering triage, ranking, act-or-wait, retry gating, research caps, and approval gates.

Watch out

Jev is one more network call and can be wrong, unavailable, or slower than the work it would skip, so keep the kill switch and safe fallback. Shadow mode enforces nothing and active mode only works if the agent honors the action. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

typesafe-skill-router

Opt-in Hermes plugin that routes each turn to one skill: a wide Jev Choice plus request checks gate a narrow fits round, injecting a single ignorable hint line.

Jev Agents & MCP#community#jevbest#catalog-import

Community

DECRUX9812/typesafe-skill-routerPython

tink-route

CLI that routes a task to one specialist skill from a Tink library using TypeSafe Jev, verifies the winner with tink mount, then prints it inline or as a mounted path.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

jon-devlapaz/tink-routePython

langchain-skill-router

LangChain deepagents middleware that replaces SkillsMiddleware with per-turn judging, loading only the SKILL.md files a turn needs so large catalogs stay out of the prompt.

Jev Agents & MCP#community#jevbest#catalog-import

Community

deyna256/langchain-skill-routerPython

jev-codex-router-skill

A Codex Skill that hands Jev the task summary plus the host's real model list, so Jev picks a model and reasoning effort for an independent subtask and Codex executes with it.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

455-dIAO/jev-codex-router-skillPython

jev-router

A Python CLI that asks Jev which of four coding agents should take a task, optionally waits for your approval, then opens that app or its website and pastes the prompt in.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

AABBAASS1/jev-routerPython

jevskill

Agent Skill that sends large tool output and CI logs to Jev and hands your expensive model only the few lines that matter, shipped with a stdlib-only caller, a doctor command, and a plan mode that keeps open-ended work on your LLM.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

lazniak/jevskillPython

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai