Kaka-cheaper

Kaka-cheaper / codeSee

Public

Visualize your project's feature logic as a semantic flow graph — not call graphs, not import maps. AI writes the data, you see the story.

ai-agent code-review developer-tools feature-map prompt-engineering
11
4
89% credibility
Found May 17, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

CodeSee is a visual canvas that shows the story of your software project instead of its code. When you work with AI coding assistants that write lots of code quickly, CodeSee helps you understand what the project does at a glance. The AI maintains a feature map that describes your project in human terms — big chapters (called Epics), individual features, and step-by-step flows. You open this map in an interactive viewer where you can zoom between different levels of detail, drag things around, and watch the picture update in real-time as your AI writes more code. It's like having a living architecture diagram that grows with your project, so you never lose control or understanding even when AI writes thousands of lines.

How It Works

1
💡 You discover CodeSee

You're working with an AI coding assistant and realize you need a way to understand what it's building without reading every line.

2
📦 You add CodeSee to your project

You run a simple setup that adds helpful instructions to your project, telling your AI assistant how to map out your features.

3
🤖 Your AI builds the map for you

When you ask your AI to write code, it automatically creates a feature map showing the story of what your project does.

4
🎨 You open the canvas and see your project come alive

You drag your feature map into the viewer and watch an interactive picture appear showing your project's big picture.

5
You explore at different levels
🗺️
Overview view

See the big picture with major chapters and how they connect

🎯
Features view

See all features grouped by their chapter, drag to organize

📋
Steps view

See the step-by-step flow within one feature, including branches and loops

6
🔄 You watch the picture grow

Turn on live mode and the canvas updates automatically as your AI writes more code, so you always see the current state.

🎉 You understand your project again

Instead of getting lost in thousands of AI-written lines, you see the story. You stay in control without reading every line.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is codeSee?

codeSee is a visual canvas that maps your project as a semantic flow graph -- showing what your code does, not how it's wired together. Instead of call graphs or import diagrams, you get a three-level drill-down: Epics (user stories), Features (capabilities), and Steps (individual actions). The twist: AI generates and maintains the map in a plain JSON file after every code change, so you can watch the graph grow as an AI writes code. The viewer is a React app that renders the graph with drag-and-drop layout, undo/redo, and live reload polling. TypeScript throughout, with install scripts that inject prompts into your project for any AI IDE.

Why is it gaining traction?

The core hook is "AI writes the code AND the documentation." When an AI generates 5,000 lines in minutes, reviewing it all is impossible -- but tracing a semantic flow graph is tractable. Unlike static analysis tools that reverse-engineer intent from code, codeSee gets AI to write the story explicitly, which is more accurate and requires fewer tokens. The zero lock-in story is compelling: plain JSON, git-diffable, works with any AI provider. The live reload feature (3-second polling) makes it feel like watching a diagram build itself in real-time.

Who should use this?

Teams doing heavy AI-collaborative development -- especially those using Cursor, Claude Code, or similar AI IDEs. Greenfield projects get the most value: install first, let AI build features, review the canvas between tasks. Developers who find themselves drowning in AI-generated code and losing architectural oversight will appreciate the "see the story, not the syntax" approach. Not for teams who prefer traditional documentation or don't use AI coding assistants.

Verdict

codeSee solves a real problem for AI-era development, but at 11 stars it's early-stage software with the credibility score of 0.9%. The concept is solid and the multi-language UI (Chinese/English) shows ambition, but limited community feedback means you should expect rough edges. Try the live demo before committing. If you are already deep in AI-collaborative workflows, this is worth watching -- just don't bet production code on it yet.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.