Free · Open Source · macOS

The quality gate for AI-generated code

CodeVetter reviews code from AI agents, catches bloat and hallucinated APIs, and sends findings back to the agent to fix. Locally or on GitHub PRs.

Download for macOSView Source

[ features ]

What it catches

Purpose-built for the patterns AI agents get wrong.

review

Review agent code

Runs AI-powered review on local diffs or GitHub PRs. Catches over-engineering, copy-paste, hallucinated APIs, and hardcoded secrets.

automation

Feedback loop

When review fails, findings are sent back to the agent as fix instructions. Re-reviews automatically until the code passes or max attempts hit.

orchestration

Orchestrate agents

Kanban board with agent personas. Assign tasks, track progress, manage concurrency. Plan, code, review pipeline runs autonomously.


[ how it works ]

Three steps to better agent code

1

Point at code

Open a local diff or connect a GitHub PR via Personal Access Token. CodeVetter picks up the changes automatically.

2

AI reviews

Your chosen AI provider analyzes the diff against configurable rules. Flags bloat, hallucinated APIs, missing error handling, and more.

3

Agent fixes

Review findings are sent back to the coding agent as structured instructions. The agent fixes issues and re-submits. Loop until clean.


[ privacy-first ]

Your code never leaves your machine

CodeVetter is a desktop app, not a SaaS. No accounts, no telemetry, no data collection. The only network call is to the AI provider you choose — and you control that API key.

Runs 100% offline (except AI API calls)
No CodeVetter servers — zero data collection
Open source — audit every line yourself
Use your own API key — switch providers anytime
Your Mac
API key only
AI Provider
No cloud. No SaaS. No middleman.

[ faq ]

Frequently Asked Questions

Something else on your mind?

Ask on GitHub

CodeVetter is a desktop app that reviews code generated by AI agents. It catches over-engineering, hallucinated APIs, copy-paste code, and hardcoded secrets — then sends findings back to the agent for automated fixes.

Yes, CodeVetter is completely free and open source under the MIT license. You only need an API key from your preferred AI provider (OpenAI, Anthropic, etc.).

CodeVetter runs entirely on your machine. The only network call is to the AI provider for generating the review. No data is sent to CodeVetter servers — there are none.

macOS 12+ currently. Windows and Linux support is planned.

Any provider with an OpenAI-compatible API — OpenAI, Anthropic, Google, and local models via Ollama or LM Studio.

CodeVetter is specifically designed for AI agent output. It catches the patterns agents get wrong: bloated abstractions, hallucinated package APIs, unnecessary complexity. It also creates a feedback loop — sending findings back to the agent for automated fixes.

Yes. Connect via a GitHub Personal Access Token and CodeVetter will review PRs directly. For local code, it reviews diffs against your current branch.


Ship better code with AI agents

Free, open source, runs offline. Requires macOS 12+ and an AI provider API key.