Stop Guessing. Start Engineering.
We’ve all been there: you ask an AI to refactor a function, it looks perfect, but then half your project breaks because of a hidden dependency. Or worse, you accidentally hardcode an API key that slips into your git push.
I built CodeDoc to solve this. It’s an MCP server for Cursor that transforms your sidebar into a Senior Architect.
<blockquote>
The “Why” Behind CodeDoc
</blockquote>
I started this project because I was hitting a wall. My company provided access to Claude and GPT-4, but like most professional teams, we have very specific development rules and internal programming principles.
Even with powerful prompting, the AI often failed to follow our custom standards—especially when analyzing “dirty” uncommitted files. It would give me generic solutions that didn’t match our architecture. I realized I didn’t need a better prompt; I needed a tool with Total Project Awareness. That is why I built CodeDoc.
<blockquote>
How CodeDoc Solves the AI “Context Gap”
</blockquote>
CodeDoc isn’t just a wrapper; it’s a custom MCP bridge that forces the AI to look at your code through the lens of your project’s specific constraints.
<ul>
<li>
Impact Analyzer: Ask “What happens if I change this?” and get a full map of the “Blast Radius” across your entire repo.
</li>
<li>
Architecture Scorecard: Get a 1-10 health score based on SOLID/OOPS principles with a path to a 10/10.
</li>
<li>
Security Sentinel: A “Pre-Commit” scan that finds leaked tokens and SQLi risks specifically in your uncommitted changes.
</li>
<li>
Continuous Evolution: We are shipping updates weekly. Phase 3 (Security &amp; Impact) is live, and Phase 4 (Auto-Fixing) is in development.
</li>
</ul>
<blockquote>
Why it’s different?
</blockquote>
<ul>
<li>
Contextual Enforcement: It scans your local environment to ensure custom coding principles are actually followed, not just suggested.
</li>
<li>
Dirty File Focus: It prioritizes uncommitted changes, ensuring new code matches the standards of the existing codebase.
</li>
<li>
Hardcoded Wisdom: It transforms abstract rules into actionable “Guardian” checks.
</li>
</ul>
I’m looking for early adopters to break it! Give it a spin on your most complex repo and let me know: What’s the one “Guardian” rule your project is missing?
👉 Check it out on GitHub: akshay1018/mcp-codedoc
#Programming #AI #Productivity #OpenSource #Cursor #SoftwareArchitecture #coding #MCP

Akshay Kumar

CodeDoc is an MCP server for Cursor that analyzes code changes against project-specific standards. It provides impact analysis to show dependencies affected by changes, architecture scoring based on SOLID principles, and security scanning for uncommitted files to detect leaked credentials and SQL injection risks. The tool focuses on enforcing custom coding principles by scanning local environments and prioritizing uncommitted changes to ensure new code matches existing codebase standards.

Meet CodeDoc: The "Guardian Engine" Every Cursor User Needs

I’m the creator! I built CodeDoc to enforce custom project rules that generic AI often misses. If you have a specific security rule or health check you want added, drop it below—I’ll include the best ones in next week’s update! I am currently working on this and getting new updates weekly.
Quick Install (Cursor Settings &gt; Features &gt; MCP):
<ul>
<li>
Name: CodeDoc | Type: command
</li>
<li>
Command: uvx
</li>
<li>
Args: [“–refresh”, “–from”, “git+https://github.com/akshay1018/mcp-codedoc.git”, “codedoc”]
</li>
</ul>
Try this: Ask <a href="https://app.daily.dev/codedoc" data-mention-id="gIk1BF8QUOW8ZYuukSZeh" data-mention-username="codedoc" translate="no">@codedoc</a> to “Check my uncommitted files for security leaks” or “Give me a health score for this file.”