Agents

AI-powered agents for automating tasks and handling conversations.

What are Agents?

Agents are AI assistants that can:

• Answer questions and have conversations
• Execute tasks based on instructions
• Make phone calls and handle voice interactions
• Use tools and integrations

Agent Types

Custom Agents (Subagents)

Specialized AI agents configured with custom prompts and tools for specific tasks.

• Text-based interactions
• Custom system prompts
• Access to knowledge base
• Tool integrations

Learn more about Custom Agents →

Voice Agents

AI agents that handle phone calls using natural speech.

• Outbound and inbound calls
• Natural voice synthesis (ElevenLabs)
• Phone integration (Twilio)
• Data collection during calls

Learn more about Voice Agents →

How Agents Work

User Input → Agent (with prompt + context) → AI Model → Response/Action
                      ↑
                Knowledge Base
                     Tools
                     History

1. User provides input - Text, voice, or trigger event
2. Agent processes with context - System prompt, knowledge, previous messages
3. AI model generates response - Using Claude or other LLM
4. Agent executes actions - Respond, use tools, call APIs

Using Agents in Flows

Agents can be used within flows:

Trigger → AI Transform (using agent) → Process Result → Action

Voice agents are triggered via the Voice Agent node:

Trigger → Voice Agent (outbound call) → Wait for completion → Process transcript

Agent Configuration

Key configuration options:

Best Practices

Writing System Prompts

• Be specific about the agent's role
• Define what the agent should and shouldn't do
• Provide examples of good responses
• Include relevant context

Managing Context

• Keep prompts concise
• Reference knowledge base instead of duplicating content
• Structure information clearly

Testing

• Test with various inputs
• Verify tool usage works correctly
• Check edge cases and error handling

Next: Custom Agents →