Skip to content

Latest commit

 

History

History
897 lines (637 loc) · 27.8 KB

File metadata and controls

897 lines (637 loc) · 27.8 KB

Welcome to my-ai-tools 👋

GitHub stars GitHub license PRs Welcome

Comprehensive configuration management for AI coding tools - Replicate my complete setup for Claude Code, OpenCode, Amp, Codex, Gemini CLI and CCS with custom configurations, MCP servers, skills, plugins, and commands.

📖 View Documentation Website - Interactive landing page with full documentation and search.

✨ Features

  • 🚀 One-line installer - Get started in seconds
  • 🔄 Bidirectional sync - Install configs or export your current setup
  • 🤖 Multiple AI tools - Claude Code, OpenCode, Amp, CCS, and more
  • 🔌 MCP Server integration - Context7, Sequential-thinking, qmd
  • 🎯 Custom agents & skills - Pre-configured for maximum productivity
  • 📦 Plugin support - Official and community plugins

🎬 Demo

IT Man Channel

IT Man - My AI Setup in 2026

📋 Prerequisites

  • Bun or Node.js LTS - Runtime for tools and scripts
  • Git - Version control
  • Claude Code subscription or use CCS with affordable providers (GLM, MiniMax)

🚀 Quick Start

One-Line Installer (Recommended)

Install directly without cloning the repository:

curl -fsSL https://ai-tools.itman.fyi/install.sh | bash

Security Note: Review the script before running:

curl -fsSL https://ai-tools.itman.fyi/install.sh -o install.sh
cat install.sh  # Review the script
bash install.sh

Options:

# Preview changes without making them
curl -fsSL https://ai-tools.itman.fyi/install.sh | bash -s -- --dry-run

# Backup existing configs before installing
curl -fsSL https://ai-tools.itman.fyi/install.sh | bash -s -- --backup

# Skip backup prompt
curl -fsSL https://ai-tools.itman.fyi/install.sh | bash -s -- --no-backup

Manual Installation

Clone the repository and run the installer:

git clone https://github.com/jellydn/my-ai-tools.git
cd my-ai-tools
./cli.sh

Options:

  • --dry-run - Preview changes without making them
  • --backup - Backup existing configs before installing
  • --no-backup - Skip backup prompt

🔄 Bidirectional Config Sync

Forward: Install to Home (cli.sh)

Copy configurations from this repository to your home directory (~/.claude/, ~/.config/opencode/, etc.):

./cli.sh [--dry-run] [--backup] [--no-backup]

Reverse: Generate from Home (generate.sh)

Export your current configurations back to this repository for version control:

./generate.sh [--dry-run]

Tip: Use generate.sh after customizing your local setup to save changes back to this repo.


🤖 Claude Code

Primary AI coding assistant with extensive customization.

Installation

curl -fsSL https://claude.ai/install.sh | bash

MCP Servers Setup

Automatic Setup (Recommended)

Run the setup script to configure MCP servers:

./cli.sh

The script will prompt you to install each MCP server:

  • context7 - Documentation lookup for any library
  • sequential-thinking - Multi-step reasoning for complex analysis
  • qmd - Quick Markdown Search with AI-powered knowledge management

Manual Setup

For Claude Desktop

Add to ~/.claude/mcp-servers.json:

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    },
    "sequential-thinking": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
    },
    "qmd": {
      "command": "qmd",
      "args": ["mcp"]
    }
  }
}
For Claude Code

Use the CLI (installed globally for all projects):

claude mcp add --scope user --transport stdio context7 -- npx -y @upstash/context7-mcp@latest
claude mcp add --scope user --transport stdio sequential-thinking -- npx -y @modelcontextprotocol/server-sequential-thinking
claude mcp add --scope user --transport stdio qmd -- qmd mcp

MCP Scopes:

  • --scope user (global): Available across all projects
  • --scope local (default): Only in current project directory
  • --scope project: Stored in .mcp.json for team sharing

Managing MCP Servers

# List all configured servers
claude mcp list

# Remove an MCP server
claude mcp remove context7

# Get details for a specific server
claude mcp get qmd

Knowledge Management

Replace deprecated claude-mem with qmd-based knowledge system:

  • Project-specific knowledge bases in ~/.ai-knowledges/
  • AI-powered search via qmd MCP server
  • No repository pollution
  • See qmd Knowledge Management Guide

Plugins

Installation

Install via setup script or manually:

# Official plugins
claude plugin install typescript-lsp@claude-plugins-official
claude plugin install pyright-lsp@claude-plugins-official
claude plugin install context7@claude-plugins-official
claude plugin install frontend-design@claude-plugins-official
claude plugin install learning-output-style@claude-plugins-official
claude plugin install swift-lsp@claude-plugins-official
claude plugin install lua-lsp@claude-plugins-official
claude plugin install code-simplifier@claude-plugins-official
claude plugin install rust-analyzer-lsp@claude-plugins-official
claude plugin install claude-md-management@claude-plugins-official

# Community plugins
claude plugin install plannotator@backnotprop
claude plugin install claude-hud@claude-hud
claude plugin install worktrunk@worktrunk

Plugin List

Plugin Description Source
typescript-lsp TypeScript language server Official
pyright-lsp Python language server Official
context7 Documentation lookup Official
frontend-design UI/UX design assistance Official
learning-output-style Interactive learning mode Official
swift-lsp Swift language support Official
lua-lsp Lua language support Official
code-simplifier Code simplification Official
rust-analyzer-lsp Rust language support Official
claude-md-management Markdown management Official
plannotator Plan annotation tool Community
prd Product Requirements Documents Local Marketplace
ralph PRD to JSON converter Local Marketplace
qmd-knowledge Project knowledge management Local Marketplace
codemap Parallel codebase analysis Local Marketplace
claude-hud Status line with usage monitoring Community
worktrunk Work management Community

Key Marketplace Plugins

codemap - Orchestrates parallel codebase analysis producing 7 structured documents in .planning/codebase/:

  • STACK.md - Technologies, dependencies, configuration
  • INTEGRATIONS.md - 3rd party APIs, databases, auth
  • ARCHITECTURE.md - System patterns, layers, data flow
  • STRUCTURE.md - Directory layout, key locations
  • CONVENTIONS.md - Code style, patterns, error handling
  • TESTING.md - Framework, structure, mocking, coverage
  • CONCERNS.md - Tech debt, bugs, security issues

prd - Generate Product Requirements Documents

ralph - Convert PRDs to JSON for autonomous agent execution

qmd-knowledge - Project-specific knowledge management (guide)

Hooks & Status Line

Configure in ~/.claude/settings.json:

PostToolUse Hooks

Auto-format after file edits:

{
  "hooks": {
    "PostToolUse": [
      {
        "matcher": "Write|Edit|MultiEdit",
        "hooks": [
          {
            "type": "command",
            "command": "jq -r '.tool_input.file_path' | { read file_path; if echo \"$file_path\" | grep -q '\\.(ts|tsx|js|jsx)$'; then biome check --write \"$file_path\"; fi; }"
          },
          {
            "type": "command",
            "command": "if [[ \"$( jq -r .tool_input.file_path )\" =~ \\.go$ ]]; then gofmt -w \"$( jq -r .tool_input.file_path )\"; fi"
          },
          {
            "type": "command",
            "command": "jq -r '.tool_input.file_path' | { read file_path; if echo \"$file_path\" | grep -q '\\.(md|mdx)$'; then npx prettier --write \"$file_path\"; fi; }"
          }
        ]
      }
    ]
  }
}

PreToolUse Hooks

Transform WebSearch queries:

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "WebSearch",
        "hooks": [
          {
            "type": "command",
            "command": "node \"~/.ccs/hooks/websearch-transformer.cjs\"",
            "timeout": 120
          }
        ]
      }
    ]
  }
}

Status Line

Using claude-hud plugin:

{
  "statusLine": {
    "type": "command",
    "command": "bash -c 'node \"$(ls -td ~/.claude/plugins/cache/claude-hud/claude-hud/*/ 2>/dev/null | head -1)dist/index.js\"'"
  }
}
Claude HUD Status Line

Tip: Auto-compact is disabled. Use claude-hud to monitor context usage.

Custom Commands, Agents & Skills

Custom Commands

Located in configs/claude/commands/:

  • /ccs - CCS delegation and profile management
  • /plannotator-review - Interactive code review
  • /ultrathink - Deep thinking mode

Custom Agents

Located in configs/claude/agents/:

  • ai-slop-remover - Remove AI-generated boilerplate and improve code quality

Skills

Local Marketplace Plugins - Installed by cli.sh from .claude-plugin/plugins/:

  • adr - Architecture Decision Records
  • codemap - Parallel codebase analysis producing structured documentation
  • handoffs - Create handoff plans for continuing work (provides /handoffs command)
  • pickup - Resume work from previous handoff sessions (provides /pickup command)
  • pr-review - Pull request review workflows
  • prd - Generate Product Requirements Documents
  • qmd-knowledge - Project knowledge management
  • ralph - Convert PRDs to JSON for autonomous agent execution
  • slop - AI slop detection and removal
  • tdd - Test-Driven Development workflows

Projects Built with AI

Real-world projects built using these AI tools:

Project Description Tools Used
Keybinder macOS app for managing skhd keyboard shortcuts Claude + spec-kit
SealCode VS Code extension for AI-powered code review Amp + Ralph
Ralph Autonomous AI agent loop for PRD-driven development TypeScript
AI Launcher Fast launcher for switching between AI coding assistants TypeScript
Tiny Coding Agent Minimal coding agent focused on simplicity TypeScript

📖 Learning Stories - Detailed notes on development approaches, key takeaways, and tools I've tried.

Recommended Community Skills

Official and community-maintained skill collections for specific frameworks:

Framework Skills Repository Description
Expo expo/skills Official Expo skills for React Native development. Includes app creation, building, debugging, EAS updates, and config management workflows.
Next.js vercel-labs/agent-skills Vercel's agent skills for Next.js and React development. Includes project creation, component generation, and deployment workflows.
Andrej Karpathy forrestchang/andrej-karpathy-skills Community skills inspired by Andrej Karpathy's coding principles and practices for AI-focused development workflows.
Skills Discovery vercel-labs/skills/find-skills Skill discovery helper. Search and install skills from skills.sh when users ask about capabilities. Uses npx skills find [query].

Installation:

# Clone skills to your local config directory
git clone https://github.com/expo/skills.git ~/.claude/skills/expo
git clone https://github.com/vercel-labs/agent-skills.git ~/.claude/skills/nextjs

Configuration Files

All configuration files are located in the configs/claude/ directory:

Local marketplace plugins are in .claude-plugin/plugins/.

Tips & Tricks

  • OpusPlan Mode: Use opusplan mode to plan with Opus and implement with Sonnet, then use Plannotator to review plans
  • Session Management: Disable auto-compact in settings. Monitor context usage with claude-hud. Press Ctrl+C to quit or /clear to reset between coding sessions. Create a plan with /handoffs and resume with /pickup when approaching 90% context limit on big tasks.
  • Git Worktree: Use git worktree with try CLI. For tmux users, use claude-squash to manage sessions efficiently
  • Neovim Integration: Check out tiny-nvim for a complete setup with sidekick.nvim or claudecode.nvim
  • Cost Optimization: Use CCS to switch between affordable providers:

🎨 OpenCode (Optional)

OpenAI-powered AI coding assistant. Homepage

Installation & Configuration

Installation

curl -fsSL https://opencode.ai/install | bash

Configuration

Copy configs/opencode/opencode.json to ~/.config/opencode/:

{
  "$schema": "https://opencode.ai/config.json",
  "instructions": ["~/.ai-tools/best-practices.md", "~/.ai-tools/MEMORY.md"],
  "theme": "kanagawa",
  "default_agent": "plan",
  "mcp": {
    "context7": {
      "type": "remote",
      "url": "https://mcp.context7.com/mcp",
      "enabled": true
    },
    "qmd": {
      "type": "local",
      "command": ["qmd", "mcp"],
      "enabled": true
    }
  },
  "agent": {
    "build": {
      "permission": {
        "bash": {
          "git push": "ask",
          "qmd": "allow",
          "qmd query": "allow",
          "qmd get": "allow",
          "qmd search": "allow",
          "$HOME/.config/opencode/skill/qmd-knowledge/scripts/record.sh": "allow",
          "$HOME/.claude/skills/qmd-knowledge/scripts/record.sh": "allow"
        }
      }
    }
  },
  "plugin": ["@plannotator/opencode@latest"]
}

Custom Agents

Located in configs/opencode/agent/:

  • ai-slop-remover - Remove AI-generated boilerplate
  • docs-writer - Generate documentation
  • review - Code review
  • security-audit - Security auditing

Custom Commands

Located in configs/opencode/command/:

  • plannotator-review - Interactive code review

🎯 Amp (Optional)

AI coding assistant by Modular. Homepage

Installation & Configuration

Installation

curl -fsSL https://ampcode.com/install.sh | bash

Configuration

Copy configs/amp/settings.json to ~/.config/amp/:

{
  "amp.dangerouslyAllowAll": true,
  "amp.experimental.autoHandoff": { "context": 90 },
  "amp.mcpServers": {
    "context7": {
      "url": "https://mcp.context7.com/mcp"
    },
    "chrome-devtools": {
      "command": "npx",
      "args": ["chrome-devtools-mcp@latest"]
    },
    "backlog": {
      "command": "backlog",
      "args": ["mcp", "start"]
    },
    "qmd": {
      "command": "qmd",
      "args": ["mcp"]
    }
  }
}

See configs/amp/AGENTS.md for agent guidelines.


🔄 CCS - Claude Code Switch (Optional)

Universal AI profile manager for Claude Code. Homepage | Documentation

Installation & Configuration

Installation

npm install -g @kaitranntt/ccs

What It Does

CCS lets you run Claude, Gemini, GLM, and any Anthropic-compatible API - concurrently, without conflicts.

Three Main Capabilities:

  1. Multiple Claude Accounts - Run work + personal Claude subscriptions simultaneously
  2. OAuth Providers - Gemini, Codex, Antigravity, GitHub Copilot (zero API keys needed)
  3. API Profiles - GLM, Kimi, OpenRouter, or any Anthropic-compatible API

Quick Start

  1. Open Dashboard:

    ccs config
    # Opens http://localhost:3000
  2. Configure Your Accounts via the visual dashboard:

    • Claude Accounts (work, personal, client)
    • OAuth Providers (one-click auth)
    • API Profiles (configure with your keys)
    • Health Monitor (real-time status)
  3. Start Using:

    ccs           # Default Claude session
    ccs gemini    # Gemini (OAuth)
    ccs codex     # OpenAI Codex (OAuth)
    ccs glm       # GLM (API key)
    ccs ollama    # Local Ollama

Configuration

CCS auto-creates config on install. Dashboard is the recommended way to manage settings.

Config location: ~/.ccs/config.yaml

See configs/ccs/config.yaml for example configuration.


🤖 OpenAI Codex CLI (Optional)

OpenAI's command-line coding assistant. Homepage

Installation & Configuration

Installation

npm install -g @openai/codex

Configuration

Located in configs/codex/:

Usage

# Start Codex CLI
codex

# Use with Ollama (local models)
codex --oss

# Use with a specific task
codex "Explain this code"

🔷 Google Gemini CLI (Optional)

Google's AI agent that brings the power of Gemini directly into your terminal. Homepage

Installation & Configuration

Installation

npm install -g @google/gemini-cli

Or using Homebrew (macOS/Linux):

brew install gemini-cli

Authentication

Gemini CLI supports multiple authentication methods:

Option 1: Login with Google (OAuth)

gemini
# Follow the browser authentication flow

Option 2: Gemini API Key

export GEMINI_API_KEY="YOUR_API_KEY"
gemini

Get your API key from Google AI Studio.

Configuration

Located in configs/gemini/:

  • settings.json - Main configuration with MCP servers and experimental features
  • GEMINI.md - Agent guidelines
  • AGENTS.md - Additional agent guidelines
  • agents/ - Custom agent definitions (.md format with YAML frontmatter)
    • ai-slop-remover.md - Clean up AI-generated code patterns
    • docs-writer.md - Generate comprehensive documentation
    • review.md - Code review with best practices
    • security-audit.md - Security vulnerability assessment
  • commands/ - Custom slash commands (.toml format)
    • ultrathink.toml - Deep thinking mode

Key Features

  • 🆓 Free tier: 60 requests/min and 1,000 requests/day with personal Google account
  • 🧠 Powerful models: Access to Gemini 2.5 Flash and Pro with 1M token context window
  • 🔧 Built-in tools: Google Search grounding, file operations, shell commands
  • 🔌 MCP support: Extensible via Model Context Protocol
  • 💻 Terminal-first: Designed for command-line developers

Usage

# Start Gemini CLI
gemini

# Include multiple directories
gemini --include-directories ../lib,../docs

# Use specific model
gemini -m gemini-2.5-flash

# Non-interactive mode for scripts
gemini -p "Explain the architecture of this codebase"

Custom Commands

Custom commands are stored in ~/.gemini/commands/ as TOML files. Example:

# Run the ultrathink command
/ultrathink What is the best approach to optimize this database query?

MCP Servers

Configure MCP servers in ~/.gemini/settings.json to extend functionality:

{
  "mcpServers": {
    "context7": {
      "url": "https://mcp.context7.com/mcp"
    },
    "sequential-thinking": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-sequential-thinking"]
    },
    "qmd": {
      "command": "qmd",
      "args": ["mcp"]
    }
  },
  "experimental": {
    "enableAgents": true
  }
}

Note: Custom agents in ~/.gemini/agents/ are automatically discovered when experimental.enableAgents is set to true.


🔄 AI Launcher (Optional)

Fast launcher for switching between AI coding assistants. Homepage

Installation & Configuration

Installation

curl -fsSL https://raw.githubusercontent.com/jellydn/ai-launcher/main/install.sh | sh

Configuration

Copy configs/ai-switcher/config.json to ~/.config/ai-switcher/:

Tools:

  • claude / c - Claude CLI
  • opencode / o, oc - OpenCode
  • amp / a - Amp

Templates:

  • review - Code review
  • commit / commit-zen - Commit messages
  • ac / commit-atomic - Atomic commits
  • pr / draft-pr - Pull requests
  • types - Type safety
  • test - Tests
  • docs - Documentation
  • simplify - Code simplification

🛠️ Companion Tools

Additional Tools & Integrations

Plannotator

Plannotator - Annotate plans outside the terminal for better collaboration. (GitHub)

Claude-Mem

⚠️ DEPRECATED - Use qmd Knowledge Management instead.

qmd Knowledge Skill

qmd Knowledge Skill is an experimental memory/context management system:

  • No repository pollution (external storage)
  • AI-powered semantic search
  • Multi-project support
  • Simple & reliable

See GitHub Issue #11 for details.

Claude HUD

Claude HUD - Status line monitoring for context usage, tools, agents, and todos.

# Inside Claude Code
/claude-hud:setup

Try

Try - Fresh directories for every vibe. (Interactive Demo)

Claude Squad

Claude Squad - Manage multiple AI agents in separate workspaces with isolated git worktrees.

Spec Kit

Spec Kit - Toolkit for Spec-Driven Development. (GitHub)

Backlog.md

Backlog.md - Markdown-native task manager and Kanban visualizer. (npm)

Agent Browser

agent-browser - Headless browser automation CLI for AI agents.

npx skills add vercel-labs/agent-browser

📚 Best Practices

Setup includes configs/best-practices.md with comprehensive software development guidelines:

  • Kent Beck's "Tidy First?" principles
  • Kent C. Dodds' programming wisdom
  • Testing Trophy approach
  • Performance optimization patterns

Copy the file to your preferred location and reference it in your AI tools.


📖 Resources


👤 Author

Dung Huynh


⭐ Show your support

Give a ⭐️ if this project helped you!

ko-fi


📝 Contributing

Contributions, issues and feature requests are welcome! See CONTRIBUTING.md.


Made with ❤️ by Dung Huynh