Scan your Claude Code configuration, visualise the dependency graph, catch orphan files, and track software versions — across every device and every project.

Free for individual developers. Teams get shared dashboards and invite-based collaboration.

Why MD4AI?

As your Claude Code setup grows — CLAUDE.md, skills, hooks, MCP servers, memory files, plan docs — it becomes hard to see the full picture. MD4AI scans everything and shows you:

• Dependency graph — which files reference which, interactively
• Orphan detection — config files nothing points to
• Broken references — links to files that don't exist
• Skills catalogue — project-specific, machine-wide, and marketplace plugins
• Stale files — anything untouched for 90+ days
• Environment drift — compare .env, Vercel, and GitHub Secrets
• Software versions — frameworks and tools detected vs latest releases

Quick Start

npm install -g md4ai

Requires Node.js 22 or later. Works on Windows (WSL2), Linux, and macOS.

# 1. Create an account at md4ai.com and set your API key
export MD4AI_SUPABASE_ANON_KEY="your-anon-key"

# 2. Log in
md4ai login

# 3. Link a project (get the ID from the dashboard URL)
cd /path/to/your-claude-project
md4ai link <project-id>

# 4. That's it — your scan results are live at md4ai.com

Or scan without an account:

md4ai scan --offline    # generates output/index.html

What You Get

Dependency Graph

See how your CLAUDE.md, skills, hooks, and config files connect. Search, zoom, and print wall sheets.

Orphan Detection

Find configuration files not reachable from any root. Grouped by folder with modification dates.

Skills Comparison

Every skill and plugin at a glance — machine-wide vs project-specific, with current status.

Software Versions

Track detected tool versions compared against latest stable and beta releases.

Commands

Scanning & Syncing

| Command | Description | |---------|-------------| | md4ai scan [path] | Scan a Claude project and push results to the dashboard | | md4ai scan --offline | Scan locally — generates output/index.html without pushing | | md4ai sync --all | Re-scan and sync all linked projects on this device | | md4ai link <project-id> | Link cwd to a dashboard project and run initial scan |

Analysis

| Command | Description | |---------|-------------| | md4ai simulate <prompt> | Show which files Claude would load for a given prompt | | md4ai print <title> | Generate a printable A3 wall-chart HTML | | md4ai init-manifest | Scaffold an env-manifest.md from detected .env files |

Account & Devices

| Command | Description | |---------|-------------| | md4ai login | Authenticate with email and password | | md4ai status | Show login status, linked folders, and last sync | | md4ai list-devices | List all devices and their linked projects | | md4ai mcp-watch | Monitor MCP server status (runs until Ctrl+C) |

One Command to Rule Them All

md4ai start    # or just: md4ai

Checks for updates, scans the current project, and starts MCP monitoring — all in one go.

Scan Output

Files found:   41
References:    41
Broken refs:   2
Orphans:       3
Stale files:   0
Skills:        19
Toolings:      55
Env Vars:      24 (manifest found)
Plugins:       17 (17 skills)
Data hash:     9868c4a8b50f...

Web Dashboard

All scan data syncs to md4ai.com where you can:

• Browse the dependency graph interactively
• View file contents with structure navigation
• Track environment variable drift across local, Vercel, and GitHub
• Monitor MCP server status per device
• Share projects with team members
• Compare skills across machines

Support

Questions, feedback, or bugs: [email protected]

Built by Testate Technologies Ltd · Changelog

Licence

MIT — see LICENCE for details.