AI Agents

Build intelligent conversational AI assistants

AI Agents are the intelligent conversational interfaces at the heart of ArcanFlows. They understand natural language, maintain context, and can take actions on behalf of your users.

What is an AI Agent?

An AI Agent in ArcanFlows is a customizable AI-powered assistant that:

  • Understands Context: Processes user messages using large language models (LLMs)
  • Maintains Memory: Remembers conversation history for coherent interactions
  • Accesses Knowledge: Queries your documents and data to provide accurate answers
  • Executes Actions: Uses tools and functions to perform tasks
  • Delegates Work: Hands off specialized tasks to sub-agents

Agent Components

Every agent consists of several configurable components:

1. Identity & Behavior

Define who your agent is and how it should behave:

  • Name: How the agent identifies itself
  • System Prompt: Core instructions that shape personality and capabilities
  • Welcome Message: First message shown to users

2. Model Configuration

Choose and configure the AI model:

  • Provider: OpenAI, Anthropic, Google, or Ollama
  • Model: GPT-4, Claude 3, Gemini, etc.
  • Parameters: Temperature, max tokens, top_p

3. Knowledge Base

Give your agent access to information:

  • Documents: PDFs, text files, markdown
  • Websites: Crawled web pages
  • Databases: Structured data sources
  • APIs: Real-time data connections

4. Tools & Functions

Enable your agent to take actions:

  • Built-in Tools: HTTP requests, data transformation
  • Custom Functions: Your own code and logic
  • External Integrations: Third-party services

5. Sub-Agents

Create specialized agents that your main agent can delegate to:

  • Specialist Agents: Domain-specific expertise
  • Task Agents: Focused on specific actions
  • Hierarchical Routing: Smart delegation based on intent

Use Cases

AI Agents excel in many scenarios:

Customer Support

User: I can't log into my account
Agent: I'm sorry to hear you're having trouble logging in. Let me help you:

1. First, can you tell me if you're seeing any error message?
2. Have you tried the "Forgot Password" option?

I can also check your account status if you provide your email address.

Sales Assistant

User: What's the difference between Pro and Enterprise plans?
Agent: Great question! Here's a comparison:

**Pro Plan ($49/mo)**
- 10 agents, 100K messages
- Standard integrations
- Email support

**Enterprise Plan (Custom)**
- Unlimited agents
- Custom integrations + SSO
- Dedicated account manager
- SLA guarantees

Would you like to schedule a demo to see Enterprise features in action?

Internal Knowledge Bot

User: What's our refund policy for annual subscriptions?
Agent: Based on our internal policies:

Annual subscriptions are refundable within the first 30 days.
After 30 days, we offer:
- Pro-rated credit for remaining months
- Account transfer to another user
- Downgrade to monthly billing

The full policy is documented in the Employee Handbook, Section 4.2.

Agent Architecture

Loading diagram...

Getting Started

Ready to build your first agent? Follow these guides:

  1. Creating Agents - Step-by-step agent creation
  2. System Prompts - Writing effective prompts
  3. Model Selection - Choosing the right model
  4. Knowledge Base - Adding documents and data
  5. Tools & Functions - Enabling agent actions
  6. Sub-Agents - Building agent hierarchies
  7. Deployment - Going live with your agent

Best Practices

Prompt Engineering

  • Be specific about the agent's role and boundaries
  • Include examples of ideal interactions
  • Define what the agent should NOT do
  • Test with edge cases

Knowledge Management

  • Keep documents up to date
  • Organize content logically
  • Use clear, consistent formatting
  • Monitor retrieval quality

Monitoring & Improvement

  • Review conversation logs regularly
  • Track user satisfaction metrics
  • Iterate on prompts based on feedback
  • Test changes before deploying

Next Steps