What is Jev-AI-Skill?

One agent skill plus one MCP server for Claude Code, Codex and Hermes. Jev, TypeSafe's decision model, makes the small judgments while your coding tool does the building, reviewing and shipping.

What do the four gate tools decide?

gate_triage: whether a cron run, poll, notification or chat event needs an agent at all. gate_verdict: which review findings are already closed. gate_decide: whether a question must go to the owner. gate_pick: which tool, skill or session fits.

Which AI clients are supported?

Built-in installation supports Claude Code, Codex and Hermes; the README says other clients need support for the skill instructions, Python helpers and MCP. Goal and run state live in files so a configured client can resume a run.

What does the Loop capability do?

It turns a build or code-repair goal into execution steps, runs configured checks, requests an independent review, and saves state so work can resume; completion depends on the checks and the separate review.

Does it need a TypeSafe API key?

Yes, Jev decisions require a TypeSafe API key. The --http mode with JEV_MCP_URL keeps that key on a single always-on server for a private network while Loop runs on each machine next to its projects.

How does it keep large models out of cheap turns?

Route picks a model capability tier, a relevant installed skill and whether to delegate, while the gate tools answer triage, verdict and tool-choice questions from a compact state instead of consuming a large-model turn.

What are the limits of the workflow?

It works with capability tiers rather than hard-coded providers, so each client maps tiers to the models it exposes; conversation history is not transferred between clients, and one executor should run at a time.

Jev directory

CommunityJev Agents & MCP

Jev-AI-Skill

One skill and one MCP server for Claude Code, Codex and Hermes: Jev gates large-model turns with gate_triage, gate_verdict, gate_decide and gate_pick, runs build-and-repair loops with independent review, routes model tiers and skills, and guards Git steps.

See highlights Open github.com

Stars: 3
Language: Python
License: MIT
Reading time: 2 min
Published by: Community
Author: king4s
Last verified: Oct 2026

king4s/jev-ai-skill

User repository on GitHub

View on GitHub

One AI skill + MCP server for Claude Code, Codex and Hermes: Jev (TypeSafe) gates large-model turns (event triage, review verdicts, owner questions, tool choice), runs build-and-repair loops with independent review, routes models and skills, and guards Git steps. Script-only watch mode and a shared HTTP server.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import logicrw agent-skill workflow

Works forJev for Decision gating Jev for Routing Jev for Code

Used inJev for Software & DevTools

Languages: Python
Platforms: MCP CLI
Capabilities: Open sourceRequires API key

Highlights

Gate ships four MCP tools: gate_triage, gate_verdict, gate_decide and gate_pick decide whether events, findings, questions and tool choices need a large model.
Loop turns a build or repair goal into steps, runs configured checks and requests independent review, with saved state another configured client can resume.
`--watch` runs a recurring check with no agent, silent on success and reporting a new failure once; `--http` keeps the TypeSafe key on one always-on server.
Route picks a capability tier, a skill and whether to delegate; find_code.py ranks code-search results so the model reads less.
Git recommends commit, push or pull-request steps within repository rules, tying publishing decisions to repository facts, review status and your authorization.

Watch out

Grounded on the truncated README and GitHub metadata reviewed 2026-10-15; the install section was cut off, so no install commands are documented here. Built-in installation supports Claude Code, Codex and Hermes only, Jev decisions need a TypeSafe API key, and one run should have a single active executor.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

MCP tool routing: Jev in front of the tool call

Oct 16, 2026

Tool menus can outgrow the context that reads them. The catalog's Jev-based routers and gates narrow the list: benchmarked tool search, per-tool scoring, fail-closed approvals, and WebMCP in the browser.

Jev Agents & MCP#mcp#tool-routing#agents

Read article

Approval policies for coding agents: hooks, guards, and gateways