NiazMorshed2007/jev-review on GitHub
- Stars
- 218
- Forks
- 20
- Language
- TypeScript
- License
- MIT
- Last push
- 17 Sep 2026
Read from GitHub on . Counts change daily.
Jev Review runs as a local MCP server and gives Claude Code, Codex, Cursor and OpenCode a structured quality score for the code they are writing, while they are still writing it. The coding agent stays responsible for finding what is wrong and fixing it. Jev’s job is a fast signal across independent quality dimensions such as correctness, complexity, changeability, modularity, tests and security, returned as typed score, choice and noul answers rather than a written review. There is deliberately no blended overall score, only per-dimension changes.
The intended rhythm is frequent, focused checkpoints: after a coherent slice of implementation, after a score-driven change, and before handing the work off. The first call in a session sets a baseline. From there, the agent inspects its own code, forms a guess about which dimension is weak, changes it, and rescores against the previous evaluation. The plugin sends a focused diff and context to the Jev API using the user’s own key. There is no hosted backend, database or author-run proxy, so the key stays on the developer’s machine.
How it uses Jev
import { jevResponseSchema, type JevResponse } from "./schema.js";
import type { JevQuestions } from "../evaluation/questions.js";
export const JEV_API_ENDPOINT = "https://api.typesafe.ai/v1/systemone";
export const JEV_MODEL = "jev-latest";
type FetchImplementation = typeof fetch;
type SleepImplementation = (milliseconds: number) => Promise<void>;
