This is a workshop project teaching AI-native development practices to engineers. The workshop covers multi-agent systems, context engineering, and building applications with Claude Code.
- AI Development Spectrum: Understanding vibe coding vs code assist vs AI-native development
- Context Engineering: The core skill for effective AI-assisted development
- Multi-Agent Workflows: Orchestrating parallel agents for complex tasks
- MCP Servers: Extending AI capabilities with external integrations
- Frontend: Next.js
- Backend: Convex
- AI Tooling: Claude Code CLI
- MCP: Context7 for documentation lookup
- Use Plan Mode for complex or risky changes - think first, execute later
- Use Agent Mode for well-defined tasks with autonomous execution
- When unsure, start with Plan Mode
- Input quality determines output quality
- Structure rules files well
- Provide relevant examples
- Reference existing patterns in the codebase
- Manage token limits by prioritizing relevant information
- Opus 4.5: Deep reasoning, complex architecture decisions
- Fast models: Quick refactors, small fixes
- Gemini 3.0: Visual/UI understanding
- Match the model to the task at hand
- Break down tasks effectively
- Set clear boundaries for each agent
- Review at checkpoints, not just at the end
- Stay in control - automation serves you
- Use @mentions strategically to manage context
.claude/
commands/ # Custom slash commands (/review, /test, /deploy)
mcp.json # MCP server configuration
rules.md # Project-specific rules (alternative to CLAUDE.md)
- Context7: Real-time documentation lookup during coding
- Additional servers can be configured in
.claude/mcp.json
- Initialize Next.js + Convex project
- Set up context rules
- Plan mode: outline app structure
- Agent mode: implement features
- Use Context7 MCP for docs lookup
- Code review workflow
- Iterate based on feedback