Hi everyone!
I’m excited to share CodeDoc, an MCP server I built to bridge the gap between “writing code” and “maintaining a professional repo.”
What is it? CodeDoc is a dedicated auditor that lives in your sidebar. Unlike standard AI chats where explanations are lost once the tab is closed, CodeDoc creates a permanent, version-controlled /documentation folder in your project.
How is it different?
-
Persistence: It treats documentation as code. It generates high-quality Markdown reports and saves them locally.
-
Project Scanning: It scans your entire directory (respecting .gitignore) to find files that need attention.
-
Audit-First: It’s designed for a “Pre-Commit” workflow—find bugs, concurrency issues, and security leaks before you merge.
Quick Install: Add this to your Cursor MCP settings:
Roadmap: I’m moving toward “Actionable Intelligence” next:
Check it out on GitHub: Google Search I’d love to hear your feedback on what audit rules you’d like to see next!
Phase 3 is LIVE – Meet the Guardian Engine
Hi everyone! Exactly 25 days ago, I shared CodeDoc as a documentation agent. Since then, I’ve been hard at work turning it into what I call the “Guardian Engine.” I’ve just pushed a update that moves us past documentation into Real-Time Project Intelligence.
What’s New in the Phase 3:
-
Architecture Scorecard (Phase 2 Completed): You can now ask: “Audit the health of login.tsx and optimize it.” CodeDoc gives you a 1-10 Health Score based on SOLID/OOPS principles and provides the refactored code in one go.
-
The Impact Analyzer: Feel free to Ask: “What happens if I change this function signature?” CodeDoc scans your entire project (not just the open file) to find every line that will break across your dependency chain.
-
Security Sentinel: I’ve added a “Pre-Push” security scan. It detects hardcoded secrets, API keys, and SQL injection risks. You can even ask: “Scan only my uncommitted files for secrets” to keep it lightning-fast on big repos.
-
Language Agnostic: Now fully tested and optimized for Java, TS/JS, Python, C++, and C#.
Why use it?
Standard AI chats are great for writing snippets, but they lack “Total Project Awareness.” CodeDoc bridges that gap by acting as a Senior Architect who knows how File A affects File Z.
How to update/install: If you already have it, just restart your MCP server in Cursor. If you’re new, add this:
Next up for Phase 4: Multi-file “Auto-Fixing”—the ability to not just find the radius, but fix all affected files in one click.
I’d love for you guys to try the Impact Analyzer on your complex files and let me know how it handles your dependency trees!
GitHub: https://github.com/Akshay1018/mcp-codedoc