Prompt Agent

📸 Screenshots

Here are visual examples of this section:

Prompt Agent - Worker Configuration Interface

The Prompt Agent is a conversational AI worker that processes user input through large language models to generate intelligent responses. It supports tool integration, web search capabilities, agent handoffs, and maintains conversation history. The agent can execute custom tools and seamlessly transfer conversations to specialized handoff agents when needed.

2. Configuration Parameters

The Worker accepts the following parameters:

model: The primary language model to use for generating responses (defaults to "openai/gpt-4.1")
fallbackModel: Optional backup model to use if the primary model fails
searchTheWeb: Boolean flag to enable web search capabilities for the agent

3. Input/Output Handles

input: Input handle - accepts the user's prompt or message as a string
output: Output handle - returns the agent's generated response as a string
instructions: Input handle - accepts system instructions to guide the agent's behavior
history: Input handle - accepts conversation history for context continuity
handoff: Output handle - connects to handoff agents for specialized task delegation
tool: Output handle - connects to custom tools the agent can execute

4. Usage Examples with Code

// Configure a Prompt Agent with web search enabled
const promptAgent = {
  type: "promptAgent",
  parameters: {
    model: "openai/gpt-4.1",
    fallbackModel: "openai/gpt-3.5-turbo",
    searchTheWeb: true
  },
  fields: {
    input: "What's the latest news about AI developments?",
    instructions: "You are a helpful AI assistant focused on providing accurate information.",
    history: previousConversation
  }
}

5. Integration Examples

The Prompt Agent serves as a central hub in conversational workflows, connecting to chat history workers for persistence, handoff agents for specialized tasks, and custom tools for extended functionality. It seamlessly integrates with web search and maintains context across multi-turn conversations.

6. Best Practices

Provide clear, specific instructions to guide the agent's behavior and response style
Configure a fallback model to ensure reliability when the primary model is unavailable
Use handoff agents for specialized tasks that require domain-specific expertise
Enable web search only when current information retrieval is necessary to optimize token usage

7. Troubleshooting Tips

If the agent fails to respond, check that the selected model is available and API keys are properly configured
Ensure handoff agents have both a model and handoff description configured to prevent execution errors
Monitor token usage through the built-in tracking to optimize model selection and conversation length
Verify tool connections are properly established if custom functionality isn't working as expected