# How Mike McQuaid actually uses coding agents

Homebrew maintainer since 2009

**If you take one thing:** `AgentIDE`, his own vibe-built agent IDE. he replaced Superset, Fork and Zed with a native app he built for sandboxed agents in parallel worktrees, now in Homebrew.

- **Agent:** Codex (cli) ([source, 2026-04-14](https://mikemcquaid.com/sandboxed-agent-worktrees-my-coding-and-ai-setup-in-2026/))
- **Agent:** Claude Code (cli) ([source, 2026-09-24](https://mikemcquaid.com/vibe-coding-developer-productivity-tools/))
- **Agent:** OpenCode (cli) ([source, 2026-09-17](https://github.com/MikeMcQuaid/dotfiles/blob/3e74ea9e29e03b55ee38d81a857351a8155d9f7a/shrc.sh#L170-L171))
- **Model:** Fable 5 ([source, 2026-09-24](https://mikemcquaid.com/vibe-coding-developer-productivity-tools/))
- **Model:** OpenAI Codex 5.4 ([source, 2026-04-14](https://mikemcquaid.com/sandboxed-agent-worktrees-my-coding-and-ai-setup-in-2026/))
- **Review:** Reviews agent work before pushing it himself; depth depends on the code ([source, 2026-04-14](https://mikemcquaid.com/sandboxed-agent-worktrees-my-coding-and-ai-setup-in-2026/))
- **Harness:** AgentIDE (his own Mac app) on sandvault, herdr and git worktrees ([source, 2026-09-24](https://mikemcquaid.com/vibe-coding-developer-productivity-tools/))
- **Rules file:** AGENTS-GLOBAL.md (19 lines), shared by Claude and Codex, kept minimal ([source, 2026-09-10](https://github.com/MikeMcQuaid/dotfiles/blob/c6f8883e71a4e5fdcc4074862acdfdec62243928/AGENTS-GLOBAL.md#L8))
- **Hooks:** git pre-push hook rejecting AI attribution trailers, refuse-coding-agent guard for scripts ([source, 2026-09-10](https://github.com/MikeMcQuaid/dotfiles/blob/a340d0ef58084694861946d587a0832a80782898/git-hooks/pre-push#L51))

## In their words

> My daily driver of choice is OpenAI Codex but I run out of tokens more quickly so have learned to be fine with either.

> By the end of a weekend with Claude Max and Anthropic's Fable 5 model and ~4 hours interacting with my computer I had something that sort of worked.

> # Slower local coding model; optimal for OpenCode plan/review and harder
# debugging.

> By the end of a weekend with Claude Max and Anthropic's Fable 5 model and ~4 hours interacting with my computer I had something that sort of worked.

> OpenAI Codex (5.4 `xhigh`): takes a while, generally does things mostly right first time, doesn't need much prompting/steering


Archetype (our label): Toolsmith. Newest source: 2026-09-24.
Card: https://www.prehistoriq.com/stacks/mikemcquaid/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/mikemcquaid.json