AG-UI Mastra Workshop
A comprehensive workshop demonstrating AG-UI (Agent User Interaction) protocol with Mastra integration. This workshop shows how to build sophisticated AI applications with shared state, multiple client interfaces, and rich user interactions.
What is AG-UI?
AG-UI is a protocol for communicating between AI Agents and Users, enabling:
- Shared-State: Real-time synchronization between agents and UI components
- Multiple Clients: Build web apps, CLI tools, mobile apps - all connected to the same agent
- Generative UI: Agents can render dynamic interface components
- Tool Integration: Seamless integration of agent tools with user interfaces
Learn more: @ag-ui/mastra on npm
Workshop Structure
This workshop is organized into 3 progressive steps, each building on the previous to demonstrate different aspects of AG-UI:
🎯 Step 1: Basic AG-UI Integration
Branch: git checkout step-1
Link: https://github.com/CopilotKit/CopilotKit/tree/main/examples/canvas/mastra-pm
Focus: Core concepts and simple state management
- Simple agent state (proverbs array)
- Basic CopilotKit integration with Mastra
- Frontend actions and generative UI
- CLI and Web clients with same agent
🎯 Step 2: Complex State & Agent Behavior
Branch: git checkout step-2
Link: https://github.com/CopilotKit/CopilotKit/tree/main/examples/canvas/mastra-pm
Focus: Structured data and agent personas
- Complex state schemas with Zod validation
- Product manager agent with specific instructions
- Working memory with structured data types
- Enhanced CLI debugging with state snapshots
🎯 Step 3: Production-Ready Application
Branch: git checkout step-3
Link: https://github.com/CopilotKit/CopilotKit/tree/main/examples/canvas/mastra-pm
Focus: Full-featured project management interface
- Complete kanban board and team management UI
- Rich React component architecture
- Professional project management interface
- Multiple interaction patterns (modals, drag-drop, etc.)
Key Learning Outcomes
By completing this workshop, you'll understand:
- ✅ Multi-Client Architecture: How to build CLI and web interfaces for the same agent
- ✅ Shared-State Management: How state synchronizes between agents and multiple UI clients
- ✅ Agent Design: Creating agents with personas, tools, and memory
- ✅ UI Integration: Building rich interfaces that react to agent state
- ✅ Production Patterns: Scalable architecture for real-world applications
Quick Start
Prerequisites
- Node.js 18+
- OpenAI API key
- Package manager (pnpm recommended)
Setup
- Clone and install dependencies:
git clone <repository-url>
cd mastra-pm-canvas
pnpm install
- Add your OpenAI API key:
echo "OPENAI_API_KEY=your-key-here" >> .env
- Choose your starting point:
# Start from the beginning
git checkout step-1
# Or jump to a specific step
git checkout step-2
git checkout step-3
Running the Workshop
Each step provides two different client interfaces for the same agent:
🌐 Web Interface (CopilotKit + React)
pnpm dev
# Opens http://localhost:3000
💻 CLI Interface (Terminal-based)
pnpm cli
# Interactive chat in your terminal
Key Point: Both interfaces connect to the same agent and share the same state. This demonstrates AG-UI's power in enabling multiple client types.
Workshop Navigation
Step-by-Step Progression
git checkout step-1 # Basic concepts
# Work through step-1, then:
git checkout step-2 # Enhanced state management
# Work through step-2, then:
git checkout step-3 # Full application
Compare Between Steps
# See what changed between steps
git diff step-1 step-2 --name-only
git diff step-2 step-3 --name-only
Architecture Overview
┌─────────────────┐ ┌──────────────────┐ ┌─────────────────┐
│ Web Client │ │ Mastra Agent │ │ CLI Client │
│ (CopilotKit) │◄──►│ + AG-UI │◄──►│ (Terminal) │
│ │ │ │ │ │
└─────────────────┘ └──────────────────┘ └─────────────────┘
│ │ │
└────────────────────────┼────────────────────────┘
│
┌──────────────┐
│ Shared-State │
│ + Memory │
└──────────────┘
Technologies Used
- Mastra: AI agent framework
- AG-UI: Agent User Interaction protocol
- CopilotKit: React AI interface components
- Next.js: React framework
- Zod: Schema validation
- LibSQL: SQLite-compatible database
Support & Resources
Next Steps
After completing this workshop:
- Experiment with custom tools and agent instructions
- Try building additional client interfaces (mobile, desktop)
- Explore more complex agent behaviors and state schemas
- Build your own production AG-UI application
Happy Building! 🚀
