Skip to main content

Decorators Reference

Agent Forge uses TypeScript decorators to provide a clean, declarative way to configure agents, tools, and framework features. This section provides comprehensive documentation for all available decorators.

Categories​

Core Decorators​

Essential decorators for basic Agent Forge functionality:

  • @agent - Configure agent properties and behavior
  • @llmProvider - Set LLM provider and configuration
  • @forge - Create AgentForge instances with automatic setup

Tool Decorators​

Decorators for integrating tools with agents:

  • @tool - Add tools to agent classes
  • @MCP - Integrate Model Context Protocol tools

A2A Decorators​

Agent-to-Agent protocol decorators for distributed systems:

RAG Decorators​

Retrieval-Augmented Generation integration:

Utility Decorators​

Performance and development utilities:

Plugin Decorators​

Plugin system integration:

  • @plugin - Register plugins with decorated classes

Usage Patterns​

Basic Agent Setup​

@llmProvider("openai", { apiKey: process.env.OPENAI_API_KEY })
@agent({
  name: "MyAgent",
  role: "Assistant", 
  description: "A helpful assistant",
  objective: "Help users with their tasks",
  model: "gpt-4",
  temperature: 0.7
})
class MyAgent extends Agent {}

Team with Tools and Visualization​

@tool(WebSearchTool)
@Visualizer()
@RateLimiter({ rateLimitPerSecond: 2 })
@llmProvider("openai", { apiKey: process.env.OPENAI_API_KEY })
@forge()
class MyTeam {
  static forge: AgentForge;
  
  static async run() {
    // Team implementation
  }
}

Distributed Agent Network​

// Server
@a2aServer({ port: 3001 })
@llmProvider("openai", { apiKey: process.env.OPENAI_API_KEY })
@agent({ name: "ServerAgent", ... })
class ServerAgent extends Agent {}

// Client
@a2aClient({ serverUrl: "http://localhost:3001/a2a" })
class RemoteAgent extends Agent {}

TypeScript Configuration​

Agent Forge decorators require proper TypeScript configuration:

{
  "compilerOptions": {
    "experimentalDecorators": true,
    "emitDecoratorMetadata": true,
    "target": "ES2020",
    "strict": true
  }
}

Decorator Order​

When using multiple decorators, order matters:

  1. @plugin - Register plugins first
  2. @Visualizer - Enable visualization features
  3. @RateLimiter - Apply rate limiting
  4. @tool/@MCP - Add tools
  5. @RAGChromaDb - Add RAG capabilities
  6. @a2aServer/@a2aClient - A2A protocol setup
  7. @llmProvider - Set LLM provider (required for most decorators)
  8. @agent - Configure agent (for Agent classes)
  9. @forge - Create AgentForge instance (last)

Error Handling​

Common decorator-related errors and solutions:

  • "LLM provider must be set" - Add @llmProvider before other decorators
  • "Can only be applied to classes extending Agent" - Ensure your class extends the Agent base class
  • "Decorator order" - Follow the recommended decorator order above
  • "experimentalDecorators not enabled" - Update your TypeScript configuration

Best Practices​

  1. Consistent Ordering - Always follow the same decorator order across your project
  2. Environment Variables - Use environment variables for API keys and configuration
  3. Type Safety - Leverage TypeScript's type checking with proper configurations
  4. Error Boundaries - Implement proper error handling around decorated classes
  5. Testing - Test decorated classes with proper mock implementations