![\"claude-control](\"dashboard/public/favicon.svg\")
\n\n \n
\n ![\"CI\"](\"https://github.com/code-by-gunnar/claude-control/actions/workflows/ci.yml/badge.svg\"){kind=icon}
\n ![\"npm](\"https://img.shields.io/npm/v/claude-control\")
\n CLI + web dashboard for discovering, understanding, and managing Claude Code configuration across all scope levels.\n
\n\n\n\n## Why\n\nClaude Code is far more capable than most users realise. Beyond the basics, it supports\ncustom agents, skills, hooks, MCP servers, plugin marketplaces, memory files with import\nchains, granular permission rules, and a four-level configuration hierarchy that merges\nsettings from managed, user, project, and local scopes.\n\nMost of this is invisible by default.\n\nConfiguration lives across dozens of files scattered through `~/.claude/`, `.claude/`,\n`.claude.local/`, and system-wide managed paths. There's no built-in overview. No way to\nsee what's active, what's overriding what, or what you're not using yet. Discovering that\nagents even exist means reading the docs. Figuring out which of your MCP servers is actually\nloading means opening three different JSON files and resolving the merge rules in your head.\n\nThis creates two problems at once: a **visibility problem** for power users who know what\nthey have but can't see the full resolved state, and a **discovery problem** for everyone\nelse who doesn't know what Claude Code is capable of in the first place.\n\n`claude-ctl` solves both. It scans your entire Claude Code configuration across every scope\nlevel, resolves the full override hierarchy, and presents everything in one place — a CLI\nand web dashboard that shows you exactly what Claude Code sees, what's configured, what's\navailable, and what's missing. The health score tells you how complete your setup is. The\nindividual pages let you explore each feature area in detail.\n\nIt's the control centre that makes Claude Code's full capability visible and understandable.\n\n## Why Not Just Open the Files?\n\nYou can — but there's a catch: Claude Code's configuration isn't what's in any single file.\nIt's the *result of merging* up to four layers together, where higher scopes silently override\nlower ones. A `deny` rule in your user settings will suppress an `allow` rule in your project\nsettings, and you'd never know by looking at either file alone.\n\n`claude-ctl` resolves the full override chain and shows you the **effective** state — what\nClaude Code actually sees at runtime. It also surfaces things that are easy to miss when\nbrowsing files: which CLAUDE.md files are being imported by other CLAUDE.md files, which MCP\nserver env vars are set vs. missing, which permissions are being overridden at which scope,\nand whether your overall setup matches recommended best practices.\n\nIf your config is simple, opening files works fine. If you're running multiple MCP servers,\ncustom agents, marketplace plugins, and project-level overrides, `claude-ctl` saves you from\nholding the whole picture in your head.\n\n## How It Fits In\n\nThe Claude Code ecosystem has grown quickly. Here's how `claude-control` relates to other\ntools you might encounter:\n\n| Tool | What it does |\n|------|-------------|\n| **claude-control** *(this)* | Discovery + visibility layer. Shows the full Claude Code config landscape — what's configured, what's available, and what's missing. |\n| `claude-code-templates` | Installs pre-built agents, commands, hooks, and MCP configs into your setup |\n| `claude-flow` | Orchestrates multi-agent workflows and swarms |\n| `FleetCode` | Runs multiple Claude Code sessions in parallel with git worktree isolation |\n| `ductor` | Controls Claude Code remotely via Telegram with cron and webhook support |\n\n**`claude-control` is intentionally read-only.** It doesn't install plugins, modify settings,\nor change Claude Code's behaviour in any way. Think of it as `git status` for your Claude\nCode configuration — a safe way to understand what's there before you make changes.\n\nIf you want to *change* your configuration, you'll do that by editing files directly or using\nthe tools above. `claude-control` helps you understand what you have and whether it's set up correctly.\n\n## What It Does\n\n`claude-control` scans Claude Code configuration files across all four scope levels (managed, user, project, local), resolves settings with full override chains, and displays the effective merged state of your setup. It shows MCP servers with automatic secret masking, extracts hooks and permissions, discovers CLAUDE.md memory files with import chains, provides configuration health scoring, and supports cross-project comparison. Everything is available as a terminal CLI or an interactive web dashboard.\n\n## Install\n\n```bash\nnpm install -g claude-control\n```\n\nThen use `claude-ctl` anywhere:\n\n```bash\nclaude-ctl scan # Discover all config files\nclaude-ctl status # Summary of configured vs missing\nclaude-ctl dashboard # Launch web dashboard\n```\n\nOr run without installing via `npx claude-control