Build an AI Agent in 10 Minutes (No Code)

Everything below is the full guide promised in the video — every prompt, copy-paste ready. No code required.

1. Install Claude Code

The foundation everything else sits on. Pick whichever surface fits your workflow — each takes under a minute to set up.

• Terminal (macOS, Linux, WSL) — the original surface: npm i -g @anthropic-ai/claude-code
• Desktop app — a native Mac & Windows window, no terminal needed
• VS Code extension — a side panel inside your editor

2. Build context

~/CLAUDE.md is the onboarding doc Claude reads before every chat. Write it once, and every session loads it automatically — your role, voice, banned words, and output defaults.

Paste this:

Help me build my CLAUDE.md from scratch. Ask about my business, voice,
banned words, output defaults. Save to ~/CLAUDE.md.

3. Build memory

Every correction becomes its own .md file. The same mistake never lands twice, and Claude gets sharper the more you use it.

The loop: you correct Claude → it saves the correction as a file → it loads that file next time.

Paste this:

From now on, whenever I correct you, save it as its own .md file at
~/.claude/memory/, indexed in MEMORY.md.

4. Build skills

One command fires a whole workflow across multiple tools — no re-typing a 200-word prompt every time you need it.

/your-skill → wires up the tools it needs → produces the output.

Paste this:

Turn this workflow into a /{name} skill. Create
~/.claude/skills/{name}/SKILL.md, wire the MCP connectors I need,
fire-able from any chat.

5. Build agents

Each agent gets one file and one job — a strategist to plan, a builder to execute, and a QA gate that fails anything below the bar (95/100).

STRATEGIST (Opus) → BUILDER (Sonnet) → QA GATE (95/100)

Paste this:

Build me an agent team for [process]. One .md per specialist at
~/.claude/agents/. Opus for analysers, Sonnet for executors.
95/100 QA gate.

6. Run on autopilot

Claude Routines turn your agent team into a cron job that runs on Anthropic's cloud. Set it once, walk away — daily at 8am, weekly on Monday, dropped straight into Notion or a folder.

Paste this:

Create a Claude Routine that runs my [skill or agent] every [schedule].
Run on Anthropic's cloud. Drop output to [folder or Notion page].

---

That's the whole loop: install → context → memory → skills → agents → autopilot. Each step takes a couple of minutes, and by the end you have an agent team that runs itself.

If this helped, drop a comment on the video and follow along for more — every guide ships with the prompts included.