Skip to main content
Agents are the core of Emanate — AI-powered assistants that handle customer conversations autonomously across voice and chat channels.

Agent Types

Voice Agents

Handle phone calls with natural conversation

Chat Agents

Website chat widgets for visitor engagement

Workflow Squads

Multi-agent orchestration for complex flows

How Agents Work

Every agent is built on four pillars:

1. Language Model

Choose from a range of AI models optimized for different use cases — general purpose, complex reasoning, multimodal, or ultra-fast responses. You can select a model in the agent builder.

2. System Prompt

The system prompt defines your agent’s behavior. Write clear instructions about the agent’s role, what it should help with, and any boundaries. See the Voice Agents or Chat Agents guides for examples.

3. Knowledge Base

Documents that provide context:
  • Product catalogs
  • FAQ documents
  • Sales materials
  • Technical specifications

4. Tools

Custom actions the agent can perform:
  • Function Tools: Connect to your business systems for inventory, pricing, and more
  • Transfer Tools: Hand off to human agents
  • SMS Tools: Send text messages
  • Calendar Tools: Book meetings

Creating an Agent

1

Choose Agent Type

Select Voice Agent, Chat Agent, or Workflow Squad
2

Configure Basics

Set name, first message, and system prompt
3

Select Model & Voice

Choose LLM provider and voice (for voice agents)
4

Add Knowledge

Upload documents for context
5

Configure Tools

Add custom actions as needed
6

Enable Lead Capture

Set up lead capture fields
7

Test & Deploy

Verify in test panel, then go live

Best Practices

System Prompts

Keep prompts focused on a specific role and list concrete capabilities. Avoid overloading the agent with too many responsibilities.

Knowledge Base

  • Start small: Begin with your FAQ and top 10 product documents
  • Keep updated: Refresh documents when products change
  • Structure clearly: Use headers and bullet points in documents

Testing

  • Test with real customer scenarios
  • Verify knowledge base queries work
  • Check lead capture extracts data correctly
  • Test edge cases (hang-ups, silence, unclear questions)

Agent Lifecycle

StatusDescription
DraftAgent being configured, not active
ActiveAgent is live and handling conversations
PausedTemporarily disabled
ArchivedNo longer in use

Next Steps

Voice Agents

Create phone-based AI agents

Chat Agents

Build website chat widgets

Knowledge Base

Manage agent knowledge

Custom Actions

Connect your business systems