From Zero to Claude Code in 19 Minutes (no code)

**TL;DR:** You don’t need any programming knowledge—just natural language to build real projects like typing games, productivity tools, and kanban apps with Claude Code. The core workflow: create a folder → switch to plan mode → answer guided questions → approve the plan → auto-build → modify via screenshots or text. ## Preparation: Account, Desktop App, and Workspace ### 1. Get a Paid Claude Account and Desktop App You need a paid Claude account (Professional plan to start, Max plan for more usage). **The desktop app is required**—the web version doesn’t have the “Code” tab. Install it like any other app. After installation, you’ll see three tabs: **Chat**, **Collaborate**, and **Code**. Click the “Code” tab to enter the build environment. ### 2. Create a Project Workspace Claude Code creates and edits files directly on your local machine, so it needs access to a folder. Recommended approach: - Create a dedicated folder (e.g., `Claude Code`), then create subfolders for each new project. - On first use, Claude will ask for folder permissions—approve them. ## Building Your First Project: A Typing Game ### Using Basic Prompts and Plan Mode Enter your request in the prompt box, for example: > “Build me a game that teaches me how to type faster. It should be visual, gamified, fun, and effective.” Then **switch to “plan mode”** (toggle above the prompt). Claude won’t build immediately; instead, it will outline a complete plan for you to adjust or approve. Send the prompt after switching to plan mode. ### Answer Guided Questions and Approve the Plan On first use, Claude will ask a few questions to clarify direction. For example: - Game style → choose “Racing game” - Learning type → “Pure arcade mode” - Visual theme → “Neon wave racing” - Typing content → “Interesting quotes and sentences” After answering, Claude returns a detailed plan including: game loop, why it works, architecture, build order, etc. You can highlight any section and modify it. When satisfied, click “Accept plan.” ### Auto-Build and Test Claude starts building. It uses a set of tools to build, test, adjust, and self-correct. For example, it once realized it used wrong parameters, fixed them, and retested. Before final delivery, it simulates opening the game, clicking buttons, and typing through levels to ensure everything works. Modifications are very intuitive: if you see any issue (like scores left behind during Claude’s testing), **take a screenshot, paste it into the prompt**, and tell Claude to delete or adjust. It will execute the change. ### Understanding Game Logic (No Code Needed) If you want to understand how the game works without triggering a build, just ask. For example: “How does this work? Does it get harder gradually?” Claude will answer without starting a build. You can then request changes like adding difficulty levels based on your understanding. ## Core Concepts: Model Selection and Persistent Memory ### Model Selection and Token Limits The dropdown menu offers many models. Rule of thumb: - **Professional plan:** Use the highest-tier model (e.g., Opus or Fable) for planning and first builds, then switch to a lower-tier model for subsequent edits to save tokens. - **Max plan:** Usually stick with the highest-tier model the whole time. ### Using `/init` to Create a `claude.md` File To give Claude persistent memory for a project, run the command **`/init`**. Claude reads all code and chat history and generates a `claude.md` file (plain text, not code) in the project folder. This file serves as a permanent reference document, covering important aspects of the project. You can ask Claude to update it at any time (e.g., “Update claude.md with all the important things we discussed”). ## Boosting Efficiency: MCPs, Plugins, and Skills ### MCPs (Connectors) MCPs connect Claude to external tools. Location: “Customize” menu in the code interface → click “+” to browse connectors. Recommended tools: - **Context 7**: Pulls up-to-date documentation for React, Expo, Tailwind, etc., so Claude doesn’t rely on old training data or outdated web info. - Gmail, Google Calendar, Notion, etc. Once connected, Claude automatically uses these tools when needed. You can disconnect at any time from the menu. ### Plugins Plugins are bundles of skills and tools. Anthropic periodically releases pre-built plugins. I recommend installing **Superpowers**, which adds multiple skills like brainstorming, sub-agents, debugging, etc. Claude will automatically invoke them in relevant tasks. ### Skills Skills are reusable workflows. Claude provides pre-built skills, and you can create your own. For example, ask Claude for a “security review” skill (find vulnerabilities, exposed API keys, etc.). To create a custom skill: after completing a workflow involving multiple prompts and iterations, tell Claude “Package this into a skill.” It will analyze your actions and generate a reusable skill. All three can be mixed: MCPs connect tools, skills provide workflows, and plugins can bundle multiple skills and MCPs. ## Practical Application: Building a Meeting Action-Item Kanban Board ### Connect Already-Used Tools (e.g., Granola) Go back to connectors, confirm the tools you need (e.g., Granola) are connected with proper permissions. ### Use a Prompt to Build the Board In a new folder, use the prompt: > “Build a kanban-style app that extracts action items from meeting notes and puts them into a to-do list, where cards can be moved between to do, in progress, and done. Use the connected Granola MCP to pull meeting notes and extract action items.” Switch to plan mode and send. After answering a few questions, Claude returns a plan. Approve and it builds. The result pulls the latest meetings, identifies all action items and assignees, and allows drag-and-drop between states. It works perfectly on the first try. ### Managing Context Window and Starting a New Session When the chat history gets long, Claude auto-compresses but might lose details. Before starting a new session, ask Claude for a summary (e.g., “Update claude.md with all the important things we discussed”). Then open a new chat, paste the summary, and point to the same folder. Claude reads `claude.md` and continues seamlessly. ## Deploying Online: Git and GitHub ### Install Git (If Not Already Installed) Ask Claude: “Check if I have Git installed. If not, install it for me.” It will give you the correct instructions for your OS. Usually pre-installed on newer Macs; Windows users need to follow manual steps that Claude guides through. ### Connect GitHub Account Let Claude connect to GitHub so it can push code and inform you of any manual steps required. The process varies by system (Mac/Windows), but it only needs to be done once. After that, your project is accessible on GitHub—ready to share or deploy to the web; GitHub also provides version control, so you can roll back to any previous state at any time. --- **Source:** From Zero to Claude Code in 19 Minutes (no code) - Futurepedia (https://www.youtube.com/watch?v=5tgHMa2Z3jc)