Agentic Coding: Build and Launch a Product with AI

Learn what coding agents can and cannot do, how they interact with files and commands, and how Claude Code, Codex, and Gemini CLI fit into a development workflow. Set up a project workspace and practice giving agents clear, bounded tasks.

• What AI Coding Agents Can and Cannot Do

  Explain three tasks a coding agent can help with and two limits that require human judgment.

• Navigate the Terminal and Understand Agent Actions

  Use the terminal to locate project files and identify commands an agent proposes to run.

• Create a Safe Project Workspace

  Create a clearly named project folder with a starter file for agent-assisted work.

• Install and Verify Claude Code, Codex CLI, and Gemini CLI

  Install and verify access to Claude Code, Codex CLI, and Gemini CLI in the project workspace.

• Project: Give an Agent a Bounded Documentation Task

  Use one coding agent to create a README from a short brief and inspect the resulting file changes.

• Choose an Agent and Set Task Boundaries

  Match a small development task to one of the three agents and write a prompt that specifies scope and limits.

• Capstone: Prepare a Product Workspace with an AI Coding Agent

  Use a coding agent to create a simple product starter workspace, then review its files and proposed commands.

Shape a product idea into user needs, core features, and a practical technical plan before asking an agent to code. Create project context and instructions that help agents understand the goal and make changes in manageable steps.

• Turn a Product Idea into a Clear User Problem

  Write a one-sentence problem statement that names a target user, their need, and why it matters.

• Describe Users and Their Needs

  Create a simple user profile and list three needs the product should address.

• Write User Stories and Success Criteria

  Write three user stories with clear criteria for deciding whether each one is complete.

• Set Up a Product Planning Workspace

  Create a GitHub repository with a Markdown project document and configure an AI coding workspace such as Cursor.

• Project: Define and Prioritize the First Product Version

  Produce an MVP brief that prioritizes user-facing features and identifies what will wait until later.

• Plan the Technical Approach and Small Build Steps

  Document a simple technical approach and divide the MVP into small, ordered tasks an AI coding agent can handle.

• Capstone: Create an Agent-Ready Project Plan

  Assemble a project context package with the product brief, user stories, technical plan, and instructions for an AI coding agent.

Direct agents to create a working application incrementally, using version control to track changes and recover safely. Review generated code, run it locally, and use clear feedback to improve features without losing control of the project.

• Set Up VS Code, Git, and an AI Coding Agent

  Install and verify VS Code, Git, and an AI coding agent for local development.

• Save a Project Checkpoint with Git

  Create a Git repository and save a working project version as a commit.

• Ask an AI Agent to Build One Small Feature

  Use a clear request to add one working feature to a starter application.

• Review, Run, and Improve Agent-Generated Code

  Inspect an agent's changes, run the application locally, and request one targeted improvement.

• Project: Build a Task Tracker with an AI Agent

  Build a local task tracker by adding features incrementally and saving Git checkpoints.

• Capstone: Improve and Safely Recover a Product Feature

  Deliver an improved application feature and restore the last working version if an agent change breaks it.

Learn how the Model Context Protocol (MCP) connects agents to external tools and data, and how reusable agent skills can guide recurring tasks. Configure these capabilities thoughtfully and assess their permissions and reliability before using them.

• Install Claude Code and Prepare a Safe Practice Project

  Install Claude Code and verify that it can open a sample product repository.

• Trace How an MCP Connection Gives an Agent Access to Tools

  Diagram how an agent uses an MCP server to request information from an external tool.

• Connect Claude Code to a Read-Only Filesystem MCP Server

  Configure the official filesystem MCP server to let Claude Code read only the sample project directory.

• Write a Reusable Claude Code Skill for Project Summaries

  Create a skill that guides Claude Code to summarize project documentation in a consistent format.

• Project: Build an MCP-Powered Product Launch Readiness Reporter

  Build a workflow that uses the filesystem MCP server and project-summary skill to produce a launch-readiness report.

• Capstone: Harden and Demonstrate a Reliable Agent Workflow

  Deliver a tested agent workflow that uses a skill and MCP server while documenting its permissions and failure checks.

Use agents to help find bugs and create tests while independently checking that the product behaves as intended. Apply beginner-friendly security practices, inspect dependencies and configuration, and prepare the application for release.

• Set Up the Local Testing Environment

  Install and verify Node.js, Git, and the project’s test tools.

• Run Existing Tests and Record Product Behavior

  Run the existing test suite and record its pass-or-fail results.

• Review Tests Created by a Coding Agent

  Review and run agent-generated tests for three important user behaviors.

• Project: Find and Fix a Product Bug

  Fix one product bug and demonstrate the correction with a passing test.

• Inspect Dependencies for Known Risks

  Run a dependency audit and identify any package that needs attention.

• Check Configuration and Protect Secrets

  Verify that secrets are kept out of tracked source files and configuration uses safe defaults.

• Capstone: Prepare the Product for Release

  Create a release checklist supported by passing tests and documented dependency and configuration checks.

Prepare the finished project for production, deploy it using a suitable hosting platform, and verify that its essential flows work online. Create a simple maintenance process for handling user feedback, fixing issues, and directing agents through future changes.

• Create a Production-Ready Launch Checklist

  Create a checklist covering the app’s essential user flows, required data, and launch risks.

• Connect Your GitHub Repository to Vercel

  Connect the project’s GitHub repository to a Vercel account and confirm Vercel can access it.

• Configure Build Settings and Environment Variables

  Configure the project’s build settings and add its required environment variables without exposing secret values.

• Deploy the Project to Production

  Deploy the project to a production URL and confirm the deployment completes successfully.

• Project: Verify Essential Flows on Your Live Product

  Test the launch checklist’s essential user flows on the live product and record any issues with steps to reproduce them.

• Capstone Project: Create a Maintenance Workflow with an AI Agent

  Create a repeatable process for turning user feedback into a tracked issue, directing an AI coding agent to fix it, and verifying the updated live product.