AgentCheck

AgentCheck

Local AI-powered code review agents for Claude Code

We believe code reviews should be faster and more efficient, focused on trade-offs, knowledge sharing and accountability, not on policing style, naming quirks, or catching trivial regressions and security issues.

Use AgentCheck to run specialized agents before you commit, directly in your dev flow. AgentCheck pulls in your real project context (docs, conventions, architecture) and local ecosystem (MCP tools, resources) to deliver focused, actionable feedback.

Highlights

- ✅ Local review — catch issues while you’re coding, not after PRs
- 🧠 Project-aware — understands your docs, conventions, and methodology
- 🔧 Open source & customizable — adapt each agent to your stack and standards
- 🖥️ Integrated flow — runs inside your environment, no black-box platforms
- 💸 Free to use — no seat licenses, no subscription fees

Quick Start

1) Install into your repo

This adds:

2) Initialize project context

Ask Claude Code to analyze your project and customize the agent templates:

This analyzes your codebase and generates project-specific context files in .claude/agentcheck/project-context/.

3) Register the task in your Claude Code notes

Add to your AGENTS.md / CLAUDE.md:

What You Get

* Diff‑aware reviews focused on what you changed
* Prioritized findings: 🔴 Critical · 🟠 High · 🟡 Medium · 🔵 Low
* Concrete fixes: suggestions you can apply immediately
* Project‑specific context: agents learn from your docs and past decisions

AgentCheck in Action

All agents running in parallel:

Comprehensive analysis results:

Agents List

Reports are written to .claude/agentcheck/reports/ and grouped by agent.

How It Works

1. Collects context: project docs, ADRs, coding standards, and your Claude Code MCPs
2.…