CodeGuard — Architecture Linter for AI Code

Check architecture consistency of AI-generated code to prevent agents from introducing bad dependencies or patterns.

Problem

With the rise of AI coding agents (like Cursor, Claude), generated code snippets may violate existing architectural conventions, such as circular dependencies, wrong layer calls, or inconsistent naming. Current linters focus on syntax and style, lacking architecture-level checks.

Solution

A configurable architecture linter that lets developers define project layers, allowed dependency directions, module boundaries, etc., then automatically scans AI-generated changes and warns about violations before commit. Runs as a CI step or editor plugin.

Users

Teams and individual developers using AI coding agents, especially tech leads maintaining architecture consistency in medium-to-large projects.

MVP

Define YAML-based architecture rules (e.g., forbid module A importing module B); parse import relationships; scan specified diffs or files and report violations; provide CLI and basic GitHub Action integration.

Revenue

Subscription per project or seat, offering team rule sync and reporting dashboard. Open-source base with paid advanced rules and enterprise integrations.

Why now

The share of AI-generated code is soaring, but architecture guarding is still manual. An indie developer can seize this 0-to-1 opportunity.