My daily battle-tested Claude Code/Desktop and OpenAI Codex setup with commands, hooks, subagents and MCP servers.

Installation • Plugins • Configuration • Statusline • References

Installation

Prerequisites: Before installing, ensure you have Claude Code and required tools installed. See INSTALL.md for complete prerequisites.

Install agents, commands, hooks, skills, and MCP servers via Claude Code Plugins system:

# Add marketplace
/plugin marketplace add fcakyon/claude-codex-settings

# Install plugins
/plugin install ultralytics-dev@fcakyon-claude-plugins # Formatting hooks + Slack/MongoDB MCPs
/plugin install github-dev@fcakyon-claude-plugins      # Git workflow agents + commands
/plugin install websearch-tools@fcakyon-claude-plugins # Tavily hooks + web search MCP
/plugin install general-dev@fcakyon-claude-plugins     # Code simplifier + utilities
/plugin install plugin-dev@fcakyon-claude-plugins      # Plugin development toolkit
/plugin install claude-tools@fcakyon-claude-plugins    # Sync CLAUDE.md + allowlist

Then create symlink for cross-tool compatibility:

ln -s CLAUDE.md AGENTS.md

Restart Claude Code to activate.

Plugins

ultralytics-dev - Auto-formatting + Slack/MongoDB/Linear MCPs

Auto-formatting hooks for Python, JavaScript, Markdown, and Bash. Includes MCP servers for Slack, MongoDB, and Linear with usage skills.

Hooks:

• format_python_docstrings.py - Google-style docstring formatter
• python_code_quality.py - Python code quality with ruff
• prettier_formatting.py - JavaScript/TypeScript/CSS/JSON
• markdown_formatting.py - Markdown formatting
• bash_formatting.py - Bash script formatting

MCPs: Slack, MongoDB, Linear

Skills:

• slack-usage - Search Slack messages and view channel history
• mongodb-usage - Query MongoDB collections and schemas

github-dev - Git workflow agents + commands

Git and GitHub automation with commit-manager and pr-manager agents plus workflow skills.

Agents:

• code-simplifier - Ensures code follows project conventions
• commit-manager - Git commit expert
• pr-manager - GitHub PR workflow automation

Commands:

• /clean-gone-branches - Clean up deleted remote branches
• /commit-staged - Commit changes with context
• /create-pr - Create pull request with context
• /update-pr-summary - Update PR description with generated summary

Hooks:

• git_commit_confirm.py - Confirmation modal before commits
• gh_pr_create_confirm.py - Confirmation modal before PR creation

Skills:

• pr-workflow - Complete PR creation workflow
• commit-workflow - Commit best practices and automation

websearch-tools - Tavily web search + hooks

Tavily MCP server for web search and content extraction with Tavily hooks and usage skill.

Hooks:

• webfetch_to_tavily_extract.py - Redirect WebFetch
• websearch_to_tavily_search.py - Redirect WebSearch
• tavily_extract_to_advanced.py - Auto-enable advanced extraction

MCPs: Tavily, Context7

Skills:

• tavily-usage - Web search and content extraction workflows

general-dev - Code simplifier + utilities

Code quality agent, architecture pattern command, and general utility hooks.

Agent:

• code-simplifier - Ensures code follows conventions

Command:

• /explain-architecture-pattern - Identify design patterns

Hooks:

• enforce_rg_over_grep.py - Suggest ripgrep
• notify.sh - OS notifications

plugin-dev - Plugin development toolkit

Complete toolkit for building Claude Code plugins with skills, agents, and validation.

Skills:

• hook-development - Create hooks with prompt-based API
• mcp-integration - Configure MCP servers
• plugin-structure - Plugin layout and auto-discovery
• plugin-settings - Per-project configuration
• command-development - Create custom commands
• agent-development - Build autonomous agents
• skill-development - Create reusable skills with progressive disclosure

Agents:

• agent-creator - AI-assisted agent generation
• plugin-validator - Validate plugin structure
• skill-reviewer - Improve skill quality

Commands:

• /plugin-dev:create-plugin - 8-phase guided plugin workflow
• /plugin-dev:load-skills - Load all plugin development skills

Hooks:

• validate_skill.py - Validates SKILL.md structure
• validate_mcp_hook_locations.py - Validates MCP/hook file locations
• validate_plugin_paths.py - Validates plugin.json paths
• validate_plugin_structure.py - Validates plugin directory structure
• sync_marketplace_to_plugins.py - Syncs marketplace.json to plugin.json

claude-tools - Sync CLAUDE.md + allowlist + context refresh

Commands for syncing CLAUDE.md and permissions allowlist from repository, plus context refresh for long conversations.

Commands:

• /load-claude-md - Refresh context with CLAUDE.md instructions
• /load-frontend-skill - Load frontend design skill from Anthropic
• /sync-claude-md - Sync CLAUDE.md from GitHub
• /sync-allowlist - Sync permissions allowlist

---

Configuration

Claude Code

Configuration in .claude/settings.json:

• Model: OpusPlan mode (plan: Opus 4.5, execute: Opus 4.5, fast: Sonnet 4.5) - source
• Environment: bash working directory, telemetry disabled, MCP output limits
• Permissions: bash commands, git operations, MCP tools
• Statusline: Custom usage tracking powered by ccusage
• Plugins: All plugins enabled

Z.ai (85% cheaper)

Configuration in .claude/settings-zai.json using Z.ai GLM models via Anthropic-compatible API:

• Main model: GLM-4.6 (dialogue, planning, coding, complex reasoning)
• Fast model: GLM-4.5-Air (file search, syntax checking)
• Cost savings: 85% cheaper than Claude 4.5 - source
• API key: Get from z.ai/model-api

Kimi K2

Run Claude Code with Kimi K2 via Anthropic-compatible API - source:

• Thinking model: kimi-k2-thinking-turbo - High-speed thinking, 256K context
• Fast model: kimi-k2-turbo-preview - Without extended thinking
• API key: Get from platform.moonshot.ai

export ANTHROPIC_BASE_URL="https://api.moonshot.ai/anthropic/"
export ANTHROPIC_API_KEY="your-moonshot-api-key"
export ANTHROPIC_MODEL=kimi-k2-thinking-turbo
export ANTHROPIC_DEFAULT_OPUS_MODEL=kimi-k2-thinking-turbo
export ANTHROPIC_DEFAULT_SONNET_MODEL=kimi-k2-thinking-turbo
export ANTHROPIC_DEFAULT_HAIKU_MODEL=kimi-k2-thinking-turbo
export CLAUDE_CODE_SUBAGENT_MODEL=kimi-k2-thinking-turbo

OpenAI Codex

Configuration in ~/.codex/config.toml:

• Model: gpt-5-codex with model_reasoning_effort set to "high"
• Provider: Azure via responses API surface
• Auth: Project-specific base URL with env_key authentication

VSCode

Settings in .vscode/settings.json:

• GitHub Copilot: Custom instructions for automated commit messages and PR descriptions
• Python: Ruff formatting with auto-save and format-on-save enabled
• Terminal: Cross-platform compatibility configurations

Statusline

The setup includes a custom statusline powered by ccusage that displays Claude usage statistics in real-time. The statusline configuration provides:

• Real-time usage tracking: Monitor token consumption and API costs as you work
• Offline support: Cached data ensures statusline works without internet connectivity
• Customizable refresh: Updates every 2 seconds for responsive feedback
• Turkish localization: Displays costs and dates in Turkish format (configurable)

Configuration

The statusline is configured through two files:

• .claude/settings.json: Contains the statusline command configuration
• .claude/ccusage.json: ccusage-specific settings for locale, timezone, and refresh behavior

For detailed setup instructions and customization options, see the ccusage statusline guide.

References

• Claude Code - Official CLI for Claude
• Anthropic Skills - Official skill examples

Thank you for the support!