The tools I use to ship software with AI agents. Grab them, try them.
Short link / QR: https://is.gd/OR4uyz
Anthropic's agentic coding tool that lives in your terminal. Understands your codebase, runs commands, edits files, handles git — all through natural language. The foundation everything else here plugs into. Available as CLI, desktop app, web (claude.ai/code), and IDE extensions (VS Code, JetBrains).
Why I use it:It's the engine — does the actual work in my repo, not just suggestions in a sidebar.
- Site:
https://claude.com/claude-code - Docs:
https://code.claude.com/docs - Install:
npm install -g @anthropic-ai/claude-code
OpenAI's agentic coding CLI. Same terminal-agent idea as Claude Code, different model family. Useful to run side-by-side and compare approaches, or as a second opinion on hard problems.
Why I use it:A second model that thinks differently — when Claude is stuck, Codex often isn't (and vice versa).
- Site:
https://openai.com/codex - GitHub:
https://github.com/openai/codex - Install:
npm install -g @openai/codex
Garry Tan's open-source Claude Code skill pack — ~30 opinionated slash-command skills that turn the agent into a virtual eng team: CEO review, design review, eng-manager review, QA, browser automation, ship/deploy. Nearly 100K GitHub stars.
Why I use it:Turns one agent into a whole team — it reviews my plans and my code before I ever look.
- Site / GitHub: https://github.com/garrytan/gstack - Skills list: https://github.com/garrytan/gstack/blob/main/docs/skills.md - Install:
git clone --single-branch --depth 1 https://github.com/garrytan/gstack.git ~/.claude/skills/gstack
cd ~/.claude/skills/gstack && ./setup
Every's plugin (Dan Shipper + Kieran Klaassen). Philosophy: each unit of work makes the next one easier. 50+ agents, 38+ skills, workflow-first: brainstorm → plan → work → review → polish → compound. 80% planning/review, 20% execution. Works with Claude Code, Codex, Cursor, and more.
Why I use it:Every task teaches the system — the agent gets sharper on my codebase instead of repeating mistakes.
- GitHub:
https://github.com/EveryInc/compound-engineering-plugin - Guide:
https://every.to/guides/compound-engineering - Install (Claude Code):
/plugin marketplace add EveryInc/compound-engineering-plugin
Mac app to run coding agents in parallel (Melty Labs). Each task gets its own git worktree — branch, files, terminal, diff, review. Spin up 5 Claudes at once, each isolated. Free app; uses your existing Claude Code / Codex login for compute.
Why I use it:Parallelism — 5 agents working different tasks at once instead of me babysitting one terminal.
AI voice-to-text for macOS / iOS / Windows. Dictate into any app — including straight into Claude Code prompts. Offline + cloud models, 100+ languages, custom AI modes. Talking is faster than typing for long prompts. Free tier; Pro / Lifetime plans.
Why I use it:I think out loud — dictating a paragraph of context beats typing it, and agents love context.
- Site:
https://superwhisper.com/ - Mac App Store:
https://apps.apple.com/us/app/superwhisper-ai-dictation/id6471464415 - Install:
brew install --cask superwhisper
(requires macOS 13+)
macOS app to browse, edit, and manage agent skills in one place (by Josh Pigford). Auto-discovers skill/agent files scattered across Claude Code, Cursor, Codex, Windsurf, Copilot, Aider, Amp. Built-in editor, live file watching, full-text search. Stop hunting through dotfiles.
Why I use it:My skills sprawl across tools and dotfiles — Chops is the one place I see and edit them all.
- Site: https://chops.md/ - GitHub: https://github.com/Shpigford/chops
Shared at the meetup — @rholdy