Skip to main content

Agents Overview

Agents are the AI-powered entities that interact with your users. Crew supports four primary agent types, each optimized for different communication channels and use cases.

Agent Types

Voice Agents

Voice agents handle phone calls with natural, low-latency speech. They can:
  • Answer inbound calls 24/7
  • Place outbound calls at scale
  • Transfer calls to humans (warm or cold)
  • Collect information through conversation
  • Trigger actions based on call outcomes
Common use cases:
  • AI receptionist for healthcare, legal, or service businesses
  • Appointment reminders and confirmations
  • Lead qualification and follow-up
  • Customer service triage

SMS Agents

SMS agents manage two-way text message conversations. They can:
  • Respond to inbound messages
  • Send outbound messages and campaigns
  • Handle multi-turn conversations with context
  • Trigger actions (appointments, notifications)
Common use cases:
  • Appointment scheduling via text
  • Order status and updates
  • Survey collection
  • Two-factor verification flows

Chat Agents

Chat agents power web-based conversational interfaces. They can:
  • Embed on any website via widget
  • Maintain session context
  • Hand off to live agents
  • Collect leads and route inquiries
Common use cases:
  • Website lead capture
  • Customer support chat
  • Product recommendations
  • Onboarding assistance

Outbound Agents

Outbound agents initiate contact rather than waiting for inbound requests. They can:
  • Place calls from uploaded lists
  • Send SMS campaigns
  • Execute multi-step outreach sequences
  • Report on campaign performance
Common use cases:
  • Appointment reminder calls
  • Payment collection outreach
  • Survey and feedback collection
  • Re-engagement campaigns

Agent Capabilities Matrix

CapabilityVoiceSMSChatOutbound
Natural language understanding
Knowledge base access
Call transfer
Appointment booking
CRM integration
Real-time transcription
Bulk campaigns
Widget embedding

Creating an Agent

1

Select Agent Type

Choose Voice, SMS, Chat, or Outbound based on your use case
2

Define Persona

Set name, voice (for voice agents), and personality traits
3

Attach Knowledge Base

Link the agent to relevant knowledge sources
4

Configure Behavior

Set guardrails, escalation triggers, and default responses
5

Assign Channel

Connect a phone number or embed code

Agent Configuration

Each agent has configurable parameters:
{
  "name": "Sarah",
  "type": "voice",
  "voice": "nova",
  "language": "en-US",
  "persona": "Friendly and professional medical receptionist",
  "knowledge_base_id": "kb_12345",
  "guardrails": {
    "max_call_duration": 600,
    "escalation_keywords": ["emergency", "urgent", "manager"],
    "pii_handling": "redact"
  }
}

Multi-Agent Crews

A single Crew can contain multiple agents with different roles:
AgentRoleChannel
SarahPrimary receptionistMain phone line
MarcusAfter-hours agentSame line, different hours
EmmaSMS responderSMS channel
Web BotLead captureWebsite widget

Best Practices

  • Distinct personas — Give each agent a unique name and personality
  • Scoped knowledge — Only attach relevant knowledge to each agent
  • Test thoroughly — Use staging agents before deploying to production
  • Monitor performance — Review transcripts and adjust behavior regularly

Next Steps