willmarple

willmarple / harness-visualizer

Public

App to help visualize agent harness artifacts

12
1
89% credibility
Found May 20, 2026 at 12 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Harness Visualizer is a web-based tool that helps people who use AI coding assistants (like Claude Code, Cursor, or Codex) understand and manage all the configuration files their assistants rely on. You point it at any project folder, and it scans for context files, skills, MCP servers, hooks, and subagents—then displays them as a colorful interactive map grouped by which AI assistant uses each piece. You can click any entry to read or edit it, and your changes save back to disk with the graph updating live. It also generates a manifest file that AI assistants can read to propose changes through the visualizer rather than editing files blindly.

How It Works

1
🔍 You discover the tool

You're using multiple AI coding assistants and want to understand what each one is actually reading and using.

2
🚀 You start it up

You run a simple command and the tool launches in your browser, ready to scan your project.

3
🗺️ Your project map appears

A colorful visual graph materializes showing every skill, rule, hook, and configuration file grouped by which AI assistant uses it.

4
👆 You click to explore

Curious about a specific skill? You click on it and the full content opens in a side panel, easy to read.

5
✏️ You make an edit

You spot something that needs updating. You edit it right there, save it, and the graph updates instantly.

6
You want to dig deeper
📊
Agents tab

See each AI assistant's complete harness with skills, hooks, and MCP servers organized in panels.

⚠️
Issues tab

Spot problems like stale files, missing descriptions on skills, or conflicting rules.

📋
Table tab

Get a flat list of everything, sortable by type, scope, or token count.

✨ You finally understand your setup

No more guessing what your AI is reading. You can see, edit, and manage your entire AI harness from one screen.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 12 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 harness-visualizer?

Harness Visualizer is a TypeScript application that gives you a real-time map of what your AI coding agents are actually reading. Point it at any project directory and it inventories every harness artifact--context files, skills, MCP server configs, hooks, and subagents--organized by which agent loads them. The frontend runs in your browser while a backend server watches your filesystem and updates the graph live. You can click any entry to inspect or edit it in place, with changes saved back to disk automatically. It also emits a derived TOML manifest that agents like Claude Code can read as a table of contents.

Why is it gaining traction?

The problem it solves is real: modern AI coding setups sprawl across a dozen folders and you lose track of what's loaded when. After a few months of adding skills and rules, you have no idea whether two agents are getting contradictory instructions. This tool answers that question in one screen. The live editing workflow is particularly clever--instead of agents editing harness files blindly, they can propose changes through the visualizer and you review them in the browser. The security model is solid too: the server binds to localhost only, validates all filesystem paths, and uses atomic writes.

Who should use this?

Developers running multiple AI coding agents (Claude Code, Cursor, Codex, etc.) who want visibility into their configuration sprawl. If you've ever wondered "wait, does Cursor actually read this file?" this answers that. Power users who edit harness files directly and want a visual inventory. Teams standardizing agent configurations across projects.

Verdict

This is a well-architected tool solving a genuine pain point, with 590 tests and a security-first design. The credibility score of 0.8999999761581421% reflects its low star count (12) and niche audience--it's production-ready in quality but early in adoption. If you manage multiple AI agents, it's worth running against your projects to see what you find. The learning curve is minimal if you just want to browse; deeper customization of pattern matching requires reading the docs.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.