What is jev-blindspot?

A prompt reviewer for Claude Code and Codex CLI that runs alongside your session. Each submitted prompt becomes a card in a local web UI showing what the request may have overlooked, such as missing context or unstated assumptions.

How does Jev decide whether to review a prompt?

A single classifier call first decides whether the prompt is worth a second look at all; only then does a lightweight model go looking, in the project directory, for what the request did not consider.

How do I install jev-blindspot?

With Node 20 or newer and a TypeSafe API key, run npm install -g jev-blindspot, write TYPESAFE_API_KEY into ~/.config/jev-blindspot/env, and run jev-blindspot install-hook. For Codex, also trust the hook once with jev-blindspot install-hook codex --trust or /hooks.

Does it block or edit my prompts?

No. It does not block your prompt, edit it, or add anything to the agent's context; your prompt goes through as written and you decide whether to act on the review.

Where do I see the reviews?

In a local web UI in a browser tab next to your session. With Claude Code Mods support it also shows as a band above the prompt in the CLI.

What does each check show and cost?

Each card shows the blind spots of the request with a pasteable sentence, and what the check cost: the model, the time, the tokens in and out, and for Claude the dollar amount and number of turns.

What are the known limits?

The README has a known-limits section that was outside the truncated digest, so this entry cannot summarise it. Requirements are Node 20+, a TypeSafe key, and Claude Code or Codex already logged in; Codex hooks additionally need a trust step.

Jev directory

CommunityJev Agents & MCP

Jev Blindspot

Prompt reviewer for Claude Code and Codex CLI: Jev first decides whether a submitted prompt needs a review, and when it does, a lightweight model flags overlooked context, unstated assumptions, or the real problem in a local web UI.

See highlights Open github.com

Stars: 12
Language: TypeScript
License: MIT
Reading time: 2 min
Published by: Community
Author: jsk4581
Last verified: Oct 2026

jsk4581/jev-blindspot

User repository on GitHub

View on GitHub

A prompt reviewer for Claude Code and Codex CLI. Finds the blind spots in each prompt you send and shows them in a local web UI, or in the CLI with the Claude Code mod.

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import logicrw prompting claude-code codex

Works forJev for Classification Jev for Decision gating

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: CLI Browser
Capabilities: Open sourceRequires API key

Highlights

Each prompt goes to Jev, which decides whether a review is needed; only then does a lightweight model look for gaps.
It does not block your prompt, edit it, or add anything to the agent's context; you decide whether to act on the review.
Install: npm install -g jev-blindspot, store TYPESAFE_API_KEY in ~/.config/jev-blindspot/env, and run jev-blindspot install-hook.
Type /blindspot in Claude Code or /prompts:blindspot in Codex; the web UI URL comes back with no model turn spent.
Each card shows a sentence you can paste into the next prompt, plus the model, time, tokens in and out, and for Claude the dollar amount.

Quickstart

bash

npm install -g jev-blindspot
mkdir -p ~/.config/jev-blindspot
echo 'TYPESAFE_API_KEY=your-key' > ~/.config/jev-blindspot/env && chmod 600 ~/.config/jev-blindspot/env
jev-blindspot install-hook

Watch out

Copy reviewed against the repository README on 2026-10-15; the README was truncated in the digest, so the sections on what leaves your machine, configuration, and known limits were not fully inspected. A TypeSafe API key is required, and the README has a what-it-costs section, so reviews spend tokens beyond your agent session.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

YouTube: Ray Amjad puts Jev inside an agentic coding loop and looks at what it costs to run.

YouTube: RoboNuggets puts Jev to work inside Claude Code.

Reddit: A Claude Code mod replaces the auto-mode classifier with Jev, 2x faster.

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