Claude Code for non-coders: the 6 parts that make it run a real job

Claude Code isn't only for developers. It reads your files, edits them, runs commands and connects to your tools, and it works inside a folder, not a chat box. Six parts turn it into something that runs a real job: the folder, CLAUDE.md, the permission mode, hooks, connectors and Mods. This video builds all six around one job every team has: onboarding a new client. Chapters 0:00 Claude Code isn't only for developers 0:50 1. What Claude Code actually is (the folder) 2:03 2. CLAUDE.md: the standing brief 3:03 3. Permission modes and deny rules 4:13 4. Hooks: the rule it can't skip 5:30 5. Connectors, skills and Mods 7:08 6. Six parts, one role The short version 1. The folder: one folder per client and a templates folder (kickoff doc, your best welcome email, the checklist). 2. CLAUDE.md: the standing brief Claude Code reads at the start of every session. Run /init for a first draft, then add what it can't work out. Short and specific beats long and vague. Anthropic's docs say Claude treats it as context, not enforced configuration. 3. The permission mode: Manual asks before edits and commands, accept edits changes files on its own, plan mode reads and proposes, auto mode runs everything with a second model checking each action. Start in plan mode; move to auto when the plan is right three times in a row. Deny rules block in every mode. 4. Hooks: a script Claude Code runs at a fixed moment. A hook that runs before an action can block it (exit code 2), and the reason goes back to Claude. CLAUDE.md asks; a hook enforces. 5. Connectors (MCP) reach the tools the job touches; add only the ones it needs. Package the steps as a skill. 6. Mods: a plugin that changes how Claude Code looks and behaves, from inside (panels, buttons, holding an action to ask you). A mod runs with your permissions and isn't sandboxed: install only from people you trust, or read your own first. Build it yourself: one job, one folder, a CLAUDE.md, plan mode until it's right three times, one hook for the mistake you can't afford, then auto mode. Tools mentioned: Claude Code (terminal, code editor, Claude desktop app, browser), MCP, Claude Code skills and Mods. Features change weekly; check Anthropic's docs. Sources (checked 2026-10-10) Claude Code overview: https://code.claude.com/docs/en/overview How Claude remembers your project (CLAUDE.md, /init): https://code.claude.com/docs/en/memory Choose a permission mode: https://code.claude.com/docs/en/permi... Automate actions with hooks: https://code.claude.com/docs/en/hooks... Hooks reference (exit code 2 per event): https://code.claude.com/docs/en/hooks Mods overview: https://code.claude.com/docs/en/plugi... Google Trends, YouTube search, US, last 30 days (read 2026-10-10): "claude code" vs "ai agent" Watch next: Claude Skills aren't enough: the 5-part system to hand a real job to AI (Day 5 link, added once it is live) AutoFlows builds AI systems on Claude around the busiest role in a company. New practical AI how-to every day: subscribe.