clauder

> CLAUDER - a safer & supercharged toolkit for Claude Code

<p align="left"> <a href="https://github.com/spacehendrix/clauder/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/github/license/spacehendrix/clauder.svg?color=00bf48"></a> <a href="https://discord.gg/7g9SrEc5yT"><img alt="Discord" src="https://img.shields.io/badge/Join-Discord-7289DA?logo=discord&logoColor=white&color=4911ff"></a> </p>

[!TIP] Would you rather chat with our docs? 💬

Note: AI responses may include mistakes. Refer to the documentation below for sample code.

Clauder is a supercharged toolkit and a safety-first configuration for Claude Code, designed to provide a robust foundation for AI-assisted development — while remaining easily configurable and extensible to any project or workflow.

It auto-updates so you always get the latest tools, and never have to think about it again.

[!WARNING] While clauder helps setting guardrails, these are insufficient to autonomously ensure correctness and safety. clauder is solely meant as a safety net and toolset, and assumes co-supervision by a human in the loop.

spacehendrix https://spacehendrix.com

Overview

This repository contains a comprehensive Claude Code configuration that provides advanced toolkits, safety mechanisms, logging, and best practices for AI-assisted development. Clauder includes:

🔒 Security & Safety

• Multi-layered secret detection and leak prevention
• File protection with immutable and ignore patterns
• Human-in-the-loop approval for sensitive operations
• Git protection against destructive operations
• Environment variable and sensitive data protection

🔎 Logging & Monitoring

• Comprehensive audit trail with SQLite database
• Real-time bash command and MCP tool logging
• Web-based tracer app for live monitoring
• Pre and post-operation validation and logging

⚡ Workflow Automation

• Automatic git checkpoints before/after each interaction
• (optional) Context engineering through documentation enforcement (HISTORY.md, SPECIFICATIONS.md)
• (optional) Audio feedback on task completion

🛠️ Advanced Toolset

• Custom commands for external AI consultation (/consult)
• Sub-agent creation and management (/spawn)
• Code review automation (/review)
• Intelligent agent recruitment (/recruit)
• 65+ on-demand MCP servers, per project

🎯 Domain-Specific Expansion Packs

• 67 specialized agents across 8 domains
• Frontend Development: React, Vue, Angular, Svelte, TypeScript specialists
• Backend Development: API architects, database specialists, security experts
• Data Science: ML engineers, data scientists, visualization specialists
• AI Development: OpenAI, LangChain, RAG, and LLM security specialists
• Infrastructure: Cloud architects, DevOps engineers, SRE specialists
• Game Development: Mechanics designers, performance optimizers, audio specialists
• Desktop Development: Electron, Tauri, Flutter desktop specialists
• General Software: System architects, UX researchers, QA strategists

💡 Smart Integration

• Automated MCP server installation, on-demand, per project
• Automatic MCP tool detection and utilization
• Automated Clauder updates, while preserving custom configuration and expansion packs
• Claude configuration backups and rollback support
• Commands recommendations and active MCP servers list on start, project based
• Status line with project name, branch, and active model, for easy monitoring (optionally, API costs per session)

Requirements

<details> <summary><strong style="display: inline; cursor: pointer; margin: 0; padding: 0;">View Requirements</strong></summary>

Also see Troubleshooting sections for further debugging tips

Supported platforms:

• macOS
• linux
• windows (experimental)

Core features should work on windows as long as running a bash terminal. Some MCP servers may not work as is and may need manual setup instead. Please raise any issue encountered and we will try to help.

Supported terminals:

• bash
• zsh

Run bash to open a bash terminal if running in an unsupported termnial (e.g. shell, powershell)

Required dependencies:

• Python 3.10+
• pip
• git
• jq
• claude

Required if using MCP servers (refer to .mcp.json for details):

• npx
• uvx
• uv
• pipx
• pnpx
• docker

</details>

Quick Start

Ensure the requirements above are met

If you do not have Clauder installed on your machine (check using which clauder):

# Clone and install repository
cd ~/ && git clone https://github.com/spacehendrix/clauder.git && cd clauder && bash ./clauder_install.sh

Navigate to your project:

# Navigate to project root directory (must be a git repository)
cd replace-with-project-root-path # && git init (if not already a repository)

In your project's .gitignore, exclude the following paths for cleaner commits:

.claude 
.claude-backup
.claude-mcp-backup
.mcp.json

Activate & start clauder in your project:

# Start clauder session
clauder # (choose ** activate in project ** when prompted)

Check section below for detailed information on how to use > CLAUDER

Documentation

[!TIP] 💬 You may also chat with this page on GitHub Chat →

Note: AI responses may include mistakes. Refer to the documentation below for sample code.

<details> <summary><code style="display: inline; cursor: pointer; margin: 0; padding: 0; color: orange; background-color: transparent; font-weight: bold;">Installation</code></summary>

Installation

Ensure the requirements above are met

# Clone repository
cd ~/ && git clone https://github.com/spacehendrix/clauder.git

# Install (must be run from the `clauder` dir)
cd clauder && bash ./clauder_install.sh

</details> <details> <summary><code style="display: inline; cursor: pointer; margin: 0; padding: 0; color: orange; background-color: transparent; font-weight: bold;">Activate Clauder in your project</code></summary>

Activate Clauder in your project

[ Important ]

Activating Clauder may override any existing .claude configuration. Backups will automatically be created to save/restore your existing configurations. These can be found at ./.claude-backup/, ./.claude-mcp-backup/ for MCP configurations.

Notably, .claude/settings.json will be overriden for consistency and security purposes, upon every activation (including auto-updates). Custom settings must be defined in .claude/settings.local.json to remain persisted throughout clauder activation.

Run in your project directory:

clauder_activate

# or you may activate it from anywhere else by providing a path to the project
# clauder_activate ./project_path

# or skip this step, start clauder, and choose 'activate' when prompted

This will copy the .claude configuration to your project.

Clauder's configuration will automatically:

• Create checkpoint commits before each session
• Protect sensitive files and directories (see .claude/.ignore and .claude/.immutable)
• Log all actions for live monitoring, or auditing purposes (see .claude/logs); enabled in .claude/preferences.md
• Enforce history and specifications tracking as you interact with Claude Code (see HISTORY.md, SPECIFICATIONS.md)
• Provides general guidelines/rules to Claude (see .claude/rules.md; Never guaranteed, but does help steering it; Do not solely rely on instructions for policing or workflows)
• Provide audio feedback on completion (optional, supports mac, linux experimental; enabled in .claude/preferences.md)
• Define custom commands for advanced workflows (e.g. /consult to consult a third party model for specific tasks, /spawn to create task specific agents, /recruit to recruit relevant agents for your project and needs, /review for a general review)
  • Required MCP servers detailed below.
• Define custom agents to help the main instance achieve specific tasks

Domain specific expansion packs available (including agents, commands, hooks and configurations - see below)

</details>

⎈ How to..

Learn how to get the best of Clauder, at your own pace

<details> <summary><code style="display: inline; cursor: pointer; margin: 0; padding: 0; color: orange; background-color: transparent; font-weight: bold;">Start a Claude session</code></summary>

How to start a Claude session

[ Important ]

Opening Claude without interacting is sufficient to index and learn all secrets in the directory. Never keep secrets such as .env in the project directory.

If secrets have been indexed or read by an AI such as Claude, you should consider removing them from the project, invalidating them and renewing them. Production secrets should be stored in a secure vault, unreadable by AI. Keeping secrets out of the working directory prevents auto-indexing, but does not prevent Claude from finding ways to access them through running commands or calling tools.

Disclaimer:

Clauder will try to prevent leaking secrets, and potentially destructive, or unrecoverable actions, by detecting unsafe operations and requesting a Human in the loop, but none of it is bulletproof.

Please make sure to supervise your AI's actions as you grant it access to sensitive or critical systems. It cannot be trusted and will inadvertently make unrecoverable mistakes, which may critically impair the company and its production services. Backup your systems, and sandbox as much as possible through restrictive AI-level access control. You are responsible for your AI's actions, as you are when using any other tool, or when managing a team.

Prefer closer, slower supervision when working on root/core nodes in your project. Equally, allow faster, lower supervision for faster iterations when working on leaf nodes, or when prototyping.

# Navigate to project root directory (must be a git repository)
cd replace-with-project-root-path # && git init (if not already a repository)

In your project's .gitignore, exclude the following paths for cleaner commits:

.claude 
.claude-backup
.claude-mcp-backup
.mcp.json

Start a new clauder session:

clauder

☕ clauder includes security features, auto-updates, and configuration backups. All Claude Code arguments supported (e.g. --continue to recall the last session)

In Claude, type:

/rules

This will define the mandatory guidelines for Claude Code.

[ Tip ]

• If your project includes a HISTORY.md file at root level, clauder will enforce keeping a history of requests and actions taken, and use it to reason about the next action to take. Comprehensive history tracking may take time.
• If your project includes a SPECIFICATIONS.md file at root level, clauder will enforce keeping an updated list of specifications as it takes actions, and use it to reason about the next action to take. When writing code manually, you may ask clauder to read the git diffs and backfill the specifications file. Comprehensive specifications tracking may take time.
• Define your secret files and folders in .claude/.ignore so clauder can guard them from being read/written.
• Define your read-only files and folders in .claude/.immutable so clauder can guard them from being overwritten.
• Exclude safe configuration files and folders in .claude/.exclude_security_checks so clauder can ommit them from safety checks (e.g. secret detection).
• Check .claude/requirements.md for prerequisites, and recommended [MCP tools](