Codevetter

Codevetter — AI skill for Claude Code

AI community

Verify AI-generated code with execution evidence — deterministic, local-first verification for coding-agent changes via a macOS app, CLI, and MCP server.

How to install Codevetter

This entry records only its repository, not the path inside it, so there is no exact command to give. Open Codevetter/codevetter and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Codevetter does

Verify AI-generated code with execution evidence — deterministic, local-first verification for coding-agent changes via a macOS app, CLI, and MCP server.

Alternatives in AI

Osaurus — Own your AI. The native macOS harness for AI agents -- any model, persistent memory, autonomous execution, cry 5.1k ★
Peekaboo — Peekaboo is a macOS CLI & optional MCP server that enables AI agents to capture screenshots of applications, o 3.2k ★
Open Science — Open Science Desktop — local-first, model-agnostic AI research workbench for macOS, Windows & Linux 1.5k ★

README

CodeVetter

AI software quality workbench for agent-generated code — desktop-first, local-first, and focused on finding bugs that normal AI review misses.

**Verify AI-generated code with execution evidence, not another LLM opinion.** CodeVetter checks a coding agent's changes by actually exercising them — running builds, tests, and browser/API behavior through a deterministic Rust engine — then emits an inspectable evidence bundle and verdict you can replay. It runs locally on your Mac as a native desktop app, a CLI, and an MCP server: no hosted PR bot, no code leaving your machine, no model grading its own homework.

How it differs

	CodeVetter	Hosted AI reviewers (CodeRabbit, Greptile)	Merge-gate tools (Verdict, PatchDrill, ProofGate)	LLM-as-judge review
Verdict based on	Executed evidence — builds, tests, runtime behavior	A model reading the diff	Diff structure + declared claims	A second model's opinion
Where it runs	Your Mac — app, CLI, MCP	Vendor cloud	Your CI / GitHub Actions	Wherever you run it
Your code leaves the machine	Never	Yes	No	Depends on provider
Open source	Yes (ISC)	Mostly no	Yes	Varies

Product Direction

CodeVetter should end as a personal verification layer for AI-built software. The durable scope is:

code review
bug finding
agent-written code verification
debugging and replay
synthetic user QA for software quality
target-audience validation after executable testing
AI step-through debugging
codebase history explanation

The near-term wedge is not beating Claude, Codex, or hosted PR bots at generic review. It is a self-first workflow that makes agent output trustworthy: inspect the diff, understand the repo and prior intent, exercise the changed behavior, preserve evidence, fix one finding at a time, and re-check that the issue is gone.