30+ CLI tools with ready-to-use Claude Code skills for AI agents. Replace bloated MCP servers with efficient CLI alternatives.
AI models have been trained on billions of terminal interactions. They already know CLI tools. MCP servers add unnecessary overhead:
- 10-32x cheaper - CLI tools use a fraction of the tokens (Scalekit benchmark)
- 72% context waste - MCP servers dump entire schemas into your context window before you ask a single question
- 100% vs 72% reliability - CLI commands succeed reliably; MCP servers have a 28% failure rate
A typical MCP server (like GitHub's) ships 93 tools costing ~55,000 tokens just to initialize. The equivalent gh CLI skill? ~800 tokens.
30 tools across 8 categories, each with a SKILL.md (Claude Code skill) and install.sh:
| Category | Tools |
|---|---|
| Cloud | aws, gcloud, az |
| DevOps | gh, docker, kubectl, terraform |
| Deployment | vercel, railway, netlify, fly, firebase, wrangler |
| Database | supabase, turso, planetscale |
| Media | ffmpeg, imagemagick, blender, yt-dlp |
| Communication | stripe, resend, twilio |
| Testing | playwright, httpie |
| Utility | jq, ripgrep, ngrok, pandoc, gh-copilot, opencli |
Copy any SKILL.md into your project's .claude/skills/ directory:
# Example: Add the GitHub CLI skill
mkdir -p .claude/skills
cp tools/github/SKILL.md .claude/skills/github.mdClaude Code will automatically pick up the skill and use gh instead of the GitHub MCP server.
# Install a specific tool
./install-all.sh --tool github
# Install a category
./install-all.sh --category cloud # Installs aws, gcloud, az
# Install everything
./install-all.sh --all
# See what's available
./install-all.sh --list# Copy all skills to your project
for dir in tools/*/; do
tool=$(basename "$dir")
cp "$dir/SKILL.md" ".claude/skills/${tool}.md"
done# Copy skills to your global Claude Code config
mkdir -p ~/.claude/skills
for dir in tools/*/; do
tool=$(basename "$dir")
cp "$dir/SKILL.md" "$HOME/.claude/skills/${tool}.md"
doneEach SKILL.md follows a consistent structure:
---
name: tool-name
description: What the tool does and when to use it
---
# Tool Name
## Authentication (if applicable)
## Common Commands
## Agent Best Practices
## Example WorkflowsSkills use YAML frontmatter so Claude Code can index and discover them automatically.
The registry.json file contains metadata for all 30 tools including:
- Install commands (brew, npm, pip)
- Official documentation URLs
- Source repositories
- Categories
To add a new tool:
- Create
tools/<tool-name>/SKILL.mdwith YAML frontmatter - Create
tools/<tool-name>/install.sh - Add the tool to
registry.json - Submit a PR
MIT