Scaffold Agents (TypeScript)¶

🐍 Looking for Python? See Python Scaffold | ☕ Looking for Java? See Java Scaffold

Generate agents with meshctl scaffold

Interactive Mode (Recommended)¶

The easiest way to create an agent:

meshctl scaffold --lang typescript

This launches an interactive wizard that guides you through:

Agent name and type
Capabilities and tools
Output directory

The generated code includes placeholder tools—you'll need to edit src/index.ts to implement your logic.

CLI Mode¶

For scripting or when you know what you want:

# Basic tool agent
meshctl scaffold --name my-agent --agent-type tool --lang typescript

# LLM-powered agent
meshctl scaffold --name emotion-analyzer --agent-type llm-agent \
  --llm-selector openai --lang typescript

# LLM provider (zero-code)
meshctl scaffold --name claude-provider --agent-type llm-provider \
  --model anthropic/claude-sonnet-4-5 --lang typescript

Agent Types¶

Type	Description	Use Case
`tool`	Basic agent with `addTool()`	Services, utilities, data processing
`llm-agent`	LLM-powered agent	AI assistants, text analysis
`llm-provider`	Zero-code LLM wrapper	Expose LLM as mesh capability

Generated Files¶

my-agent/
├── src/
│   └── index.ts      # Agent code - edit this to add your logic
├── package.json      # Add your dependencies here
├── tsconfig.json     # TypeScript config (ready to use)
├── Dockerfile        # Container build (ready to use)
├── helm-values.yaml  # Kubernetes config
└── README.md

After scaffolding:

cd my-agent && npm install
Edit src/index.ts to implement your tool logic (placeholder returns "Not implemented")

Generate Docker Compose¶

# Generate docker-compose.yml for all agents in directory
meshctl scaffold --compose

# With observability stack (registry + Redis + Tempo + Grafana)
meshctl scaffold --compose --observability

Local Tracing Setup

Use --compose --observability even if you run agents locally. Start the infrastructure with docker compose up -d, then run agents with meshctl start—they auto-connect to the Docker registry. This enables meshctl call --trace and meshctl trace.

Preview Before Creating¶

# Dry run - see what would be generated
meshctl scaffold --name my-agent --agent-type tool --lang typescript --dry-run

More Options¶

# See all scaffold options
meshctl scaffold --help

# See available agent templates
meshctl scaffold --list-modes

Next Steps¶

Continue to Run Agents →