Mdeux25

Mdeux25 / agentwatch

Public

A visual debugger for AI agents — watch your agent work in a live 3D codebase map

16
1
89% credibility
Found Apr 01, 2026 at 16 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

AgentWatch is a desktop app that visualizes Claude AI working on your codebase in a 3D spatial map with live diffs, edit history, and architecture insights.

How It Works

1
🔍 Discover AgentWatch

You find this cool tool that lets you watch an AI work on your code like a movie in 3D space.

2
🛠️ Get ready

Install a few free programs on your Mac so the AI helper can join your project.

3
🚀 Start the app

Run the program and it opens a welcoming screen ready for action.

4
📁 Open your project

Pick the folder with your code, and it instantly maps everything in a colorful 3D view.

5
🧠 Chat and watch

Talk to the AI in the chat box, and see its glowing orb zip around your code map making changes live.

✨ Guide like a pro

You direct the AI, review every edit in real time, and build faster with full understanding.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 16 to 16 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 agentwatch?

AgentWatch is a visual debugger for AI agents like Claude, rendering your codebase as a live 3D treemap on agent watch github where an agent sphere zips between files during reads and edits. Built in TypeScript with Tauri for desktop, React for UI, and Three.js for spatial viz, it streams agent events into chat logs, inline diffs, edit histories, mind maps, and architecture summaries. Solves opaque AI coding by letting you supervise spatially—watch activity hotspots, token costs, and file summaries without sidebar guesswork.

Why is it gaining traction?

Unlike github visual studio code Copilot chats or intellij visual debugger steps, it flips to agent-first: Claude drives via subprocess, you observe pulsing avatars, treemaps, and health dashboards in real-time. Hooks devs with 3D navigation, gitignore-aware scans, symbol chips, and Claude-powered arch intel—zero polling, pure event-driven flow. Early adopters rave about spotting agent patterns visually over text dumps.

Who should use this?

Mac devs directing Claude Code on TS/JS repos, especially when auditing agent edits in 100-600 file projects. Ideal for AI workflow debuggers tired of blind prompts, needing spatial oversight like visual debugger vscode but for agents—not solo coders.

Verdict

Promising alpha (16 stars, 0.9% credibility) for agent supervision fans—docs solid, features polished despite macOS/Claude limits. Grab if visual github copilot vibes appeal; skip for production without multi-platform polish.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.