Modernized devcontainer setup with Claude, Codex, Antigravity, and opencode that's ready to run in YOLO mode. Pre-configured with obra/superpowers and sane defaults that persists logins, configs, and conversation history across container rebuilds. Just drop the .devcontainer/ folder into any repo and you're good to go.
I found myself fighting with a swamp of configuration and long-build times every time I wanted to spin up a devcontainer for agentic coding: installing CLIs, configuring shell, setting up volumes for persistent state (like logins, conversation history, ...), and so on. It was inertia I didn't like.
- Coding CLIs — Claude Code, Google Antigravity (
agy, gemini-cli's successor), OpenAI Codex, opencode — pre-installed with yolo-mode aliases - Superpowers — obra/superpowers skills auto-installed for Claude, Codex, and opencode (Antigravity has no superpowers plugin yet)
skillsCLI — vercel-labs/skills package manager for agent skills —skills add <pkg>works across Claude, Codex, opencode, and more- DeepWiki MCP — pre-wired for Claude and Codex for instant Q&A on any public GitHub repo
- Persistent state — named docker volume keeps your logins, configs, and conversation history across container rebuilds
- Shell — zsh + starship + (fd, ripgrep, fzf, eza, zoxide) + lazygit (
lzg) + lazydocker (lzd)
| Alias | What it runs |
|---|---|
cy |
claude --dangerously-skip-permissions |
csy |
claude --dangerously-skip-permissions --model sonnet |
ay |
agy --dangerously-skip-permissions |
xy |
codex --yolo |
-
Copy
.devcontainer/into your repo root (also works to update an existing FAFO scaffold to the latest — re-running replaces.devcontainer/, so back up any local edits first):rm -rf _fafo-tmp .devcontainer && \ git clone --depth 1 https://github.com/epicwhale/fafo-devcontainer.git _fafo-tmp && \ mv _fafo-tmp/.devcontainer . && rm -rf _fafo-tmp
-
Open in VS Code → Dev Containers: Reopen in Container
-
Start coding freely: type
cy,xy,ay, oropencode
Using devcontainer cli? devcontainer up then devcontainer exec zsh.