Set Up Claude Code in One Shot: Let the Agent Install Skills, MCPs & Plugins

Prompt Architect · 2026-06-19 · 7 min read

TL;DR — Hand one URL to Claude Code/codex and the agent reads it, then installs the skills, MCP servers and plugins that fit your project. An agent-consumable setup registry for vibe coders.

Do you re-do the same setup every time you open a new project? Copy your favorite Claude Code skills, register MCP servers one by one, install plugins… your vibe-coding flow breaks right here, every time.

So we built a setup registry that installs itself: hand one URL to your agent, and it reads the catalog and installs only what fits your project.

Setting up an AI coding agent in the terminal

The key idea: this page isn't for you to read

A typical "recommended tools" page is read by a human who installs things one by one. This one is different — the reader is your AI agent (Claude Code / codex). Because the matching intelligence lives in the agent, you don't even need to describe "your situation" — the agent is already looking at your codebase.

How to use it: one line

Paste this into Claude Code or codex:

Read https://promptarchitect.ai.kr/setup/registry.json and install the Claude Code skills/MCPs/plugins that fit my current project. Ask me for any secret values.

The agent then ① fetches the catalog, ② detects your language/framework/task, ③ picks suitable items by their tags and "use when" notes, and ④ drops skills into .claude/skills/ and registers MCPs via claude mcp add. If a key is needed, it asks you.

What's inside

  • Skills: peer-rerun (revive a stuck CLI), triad (3-model cross-check), git-push (verified push gate), e2e-test, ui-coherence, session-status, and more.
  • MCP servers: GitHub, Context7 (live docs), Playwright (browser), Filesystem, Brave Search, Memory, Fetch.
  • Plugins / settings snippets: official plugins, statusLine, and more.

Every secret is a ${PLACEHOLDER} — we never ship real keys, and you supply your own. No remote curl|bash auto-execution either: every item is an inspectable raw file, applied by the agent under your review.

Why it helps

  • No repetition: stop hand-configuring every new repo.
  • Context-aware: Python backend? It grabs fastapi-tagged items. Frontend? ui-coherence. The agent decides.
  • Always fresh: the catalog updates in one place; next time you hand over the URL, you get the latest.

Try it now

👉 Browse the human catalog: /setup · machine-readable: /setup/registry.json, /setup/llms.txt

Want sharper prompts too? Try scoring your prompt or finding your AI tool.