AGENT EXTENSIBILITY

MCP & Agent Skills

Expose a decision capability through a reusable tool contract while keeping credentials and side effects bounded.

1. The Production Bottleneck

Agent runtimes need reusable tools, but every integration adds schema, versioning, authentication and error-handling concerns. A protocol wrapper does not remove those operational responsibilities.

2. Architecture Comparison: Without vs With Jev

WITHOUT JEVTraditional Prompt Loop
✕ Ad-hoc integration contracts

Each client can invent different names, payloads and error semantics.

✕ Hidden credential boundaries

A tool that can act on behalf of a user needs a clear owner and scope.

✕ Unversioned capability drift

Clients and servers can disagree after a tool changes its schema.

WITH JEVType-Safe Decision Layer
✓ Protocol-visible tool contract

Publish names, inputs, outputs and errors through the MCP or skill surface.

✓ Local or hosted deployment choice

Keep execution near the data when needed and document the trust boundary.

✓ Decision and execution separated

A fast judgment tool can return a choice while another layer owns side effects.

3. Execution Micro-Flow

How state flows through the bounded decision layer during live agent execution:

STEP 01
Agent requests a documented capability
STEP 02
MCP or skill server evaluates
STEP 03
Structured result and error
STEP 04
Agent or application executes the next step

4. Practical Engineering Trade-offs

✓Deploy This Pattern When:

  • Several agent clients need the same bounded capability
  • Inputs and outputs can be versioned
  • The server has explicit authentication and resource limits

⚠Avoid Or Re-evaluate When:

  • Clients that do not support the required protocol behavior
  • Local file or browser access beyond the server sandbox
  • Assuming protocol compliance proves business correctness

5. Sourced Community Builds

51 Implementations
COMMUNITY OBSERVEDby Lomesh Dutta

Lomesh Dutta Skill Router

A Claude Code session-start hook uses Jev to rank installed skills against the user's goal.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Sarah Drasner

WebMCP Side Panel - Keystroke Tool Selector

A Chrome side panel that uses Jev to select and fill tools exposed by the current page through WebMCP.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Joey Kudish~4s end-to-end for that reported example$0.0016 for one reported Wikipedia navigation example

Jev Browser

A Jev-driven browser agent exposed as an MCP server, CLI, or library.

Inspect Code & Execution Trace →
OFFICIALby TypeSafe AI

TypeSafe Agent Skills

Official skills that teach coding agents how to design and compose TypeSafe decision workflows.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Joey KudishProject reports roughly 150–500ms for most judgments

Jev MCP

Eleven typed Jev judgment tools exposed through MCP for verification, screening, routing and review.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Kitze

Skillbox Tool Matcher

A self-hosted skills library with optional task-aware Jev recommendations.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Cline

jev-browser in Cline

A marketplace plugin that connects Jev-driven browser execution directly into the Cline coding agent.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Ranjan

AskJev (MCP Agent)

Claude handles conversational intent while Jev directs browser navigation across live websites.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Instructa

Switchloom

A Codex model-routing experiment whose author reports no advantage in two pilots.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby jackbarunz

Jev Tool Router

Narrows MCP tool discovery for Codex using Jev and explicit fallbacks.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Jeff Emanuel

SkillRanker (Rust CLI)

A high-performance Rust CLI where Jev ranks available agent skills for the next action.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Abhishek Ashok Kumar

Jev MCP Dispatcher

Executes plain-language MCP tool calls with zero general LLMs, driven entirely by Jev.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Kun Chen

Compact Adviser

Uses Jev to recommend whether the current agent session is safe to compact.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Konstantinos Botonakis

Codex Context Diet

A Codex plugin where Jev trims bulky tool results before they pollute the agent context window.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby heyman333

Jev CLI for Claude Code & Codex

A zero-dependency Jev CLI and skill for narrow coding-agent judgments.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby burnigtm

Jev MCP Coding Loop

A local MCP server exposing typed Jev judgments to coding agents.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby さよ☆マギカ

LogJev

An open-source Jev-style decision layer that accepts text, images, or audio and returns bounded choices or scores.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Daniel San

Claude Code Jev Model Router Mod

A native CLI mod for Claude Code that routes terminal coding tasks through direct TypeSafe API or Vercel AI Gateway.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Kai

Jev Browser Skill (jb)

A standalone Playwright Chromium CLI that uses Jev to choose in-page browser actions.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby PatCodes

Open-Source Jev? Install Laya Locally + 3 Useful Demos

Step-by-step tutorial installing the open-source 420M parameter Laya model locally on CPU via ONNX with 3 runnable demos.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Elvis

Elvis Sun Jev Skills Workflow

A Newsjack news-filtering case; the author says it processed 384 news items in 24.9 seconds for $0.19.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Ronak Malde

Rronak Jev Skills Workflow

Astra plans while Jev controls Minecraft actions, demonstrating a win against the Ender Dragon; the author reports the time and cost.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Muhammad Ayan

Social with Aayan Jev Skills Workflow

Introduces 10 Jev repositories, including browser agents and Claude Code context cleanup.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Avi Chawla

_avichawla Jev Skills Workflow

Jev is used to evaluate agent execution traces and, with the Beacon memory layer, reuse those signals across several coding agents.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Daniel San

Dani Avila Jev Skills Workflow

A skill-selection approach for Claude Code: send the skill list to Jev to judge, then inject only the matching skills.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby 超级峰

Mileson Jev Skills Workflow

A Chinese-language beginner tutorial on Jev for developers just starting to learn and try it.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Steve Darlow

Steven Darlow Jev Skills Workflow

Jev + Hermes for routing, memory, skill selection and GUI clicks; the author says about 0.4 seconds.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby 鈴木@アナログ営業会社を100日後にAIで売上を300%にする人

ai_300 Jev Skills Workflow

Relays Newsjack's comparison data: filtering 384 news items, compared with Opus.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Ruben Hassid

Ruben Hassid Jev Skills Workflow

Recommends an introductory Jev video, stressing that it lets agents decide their next step faster.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby 森叔

Harrison Jev Skills Workflow

Chinese-language setup notes: request access, install the official TypeSafe skill and configure the API key.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby K2S

K2S Jev Skills Workflow

Introduces 7 Jev skills that can be used to build agents.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Khairallah AL-Awady

Eng Khairallah Jev Skills Workflow

A 10-minute Jev getting-started guide, from access and installation to the first decision result.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Chris

Everest Chris Jev Skills Workflow

An introduction to a guide on monetizing and using Jev; the post itself mainly points to the article.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Bowe Frankema

Bowe Frankema Jev Skills Workflow

Recommends installing the official skill, then having a coding agent find places in a project where Jev fits.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby そう|Claude CodeのX運用で月800万インプ

so_ainsight Jev Skills Workflow

A collection of Japanese open-source projects covering context compaction, browser agents, search and more.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Tjay

type_tjay Jev Skills Workflow

An overview of a tool stack that puts Jev into marketing and video production; mostly promotional.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Ronin

DeRonin Jev Skills Workflow

A Jev installation guide with notes on where it fits, including coding-agent integration.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Jason Zhu

GoSailGlobal Jev Skills Workflow

A search re-ranking counterexample: Jev tested on a 33,047-entry skill directory, questioning whether re-ranking alone clearly beats vector retrieval.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby zhouluobo

Zhou Luobo Jev Skills Workflow

A Chinese-language roundup of using Jev for routing, memory, skill selection and compaction in Hermes, Claude Code and Codex.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby VFX-DESK @毎朝更新CG・AI最新ニュースサイト

VFXDESK Jev Skills Workflow

A Japanese introduction to `wuyoscar/jev-skill`, covering its skills and examples and explaining the three judgment types Choice, Noul and Score.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Niaz Morshed

JCR · Jev Capability Resolver

Turns a vague agent request into a deterministic command.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Elia Alberti

jev-rules

Jev picks which of your rules apply, so Claude only sees those.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby MUKi

Jev Search MCP

Plain-language web search where Jev picks the sources.

Inspect Code & Execution Trace →
FIRST-HAND TESTby MathBullied

jevgrep

Semantic code search for agents: find behaviour, not strings.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Safzan Pirani

pi-jev-skill-picker

Ranks the Pi agent’s skills for the task in front of it.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby タイド

typesafe-mcp

An MCP connector that gives Claude Code, Claude Desktop, and Codex typed Jev judgments.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Thin-Illustrator-912

Jev Agent Toolkit

A portable coding-agent Skill with an optional Jev MCP bridge.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby Paddy Srinivasan

DigitalOcean Action Gateway Jev Tool

Jev exposed as a governed, discoverable tool alongside DigitalOcean's large tool catalog.

Inspect Code & Execution Trace →
COMMUNITY OBSERVEDby 0xShikhar⚡️

JEV Fuse

A governed execution layer that turns Jev probabilities into deterministic policy actions.

Inspect Code & Execution Trace →
FIRST-HAND TESTby Daniel Agrici

Jev SEO Audit by Daniel Agrici

A Claude Code skill and CLI that audits a live site, keeps rule checks in code, and uses Jev for typed page judgments.

Inspect Code & Execution Trace →
SECONDARY REPORTby MIKE

Jev Skills for Routing, Guardrails and Fan-out

Reusable community skills for adding Jev decisions to agent workflows.

Inspect Code & Execution Trace →