Creating Your First AI Agent
An AI agent is a customizable chatbot with specific instructions, personality, and capabilities. This guide walks you through creating your first one.
Before You Start
You need:
- A workspace (client account) created
- Editor or Admin access to the workspace
- At least one AI model enabled by your agency
Step-by-Step Creation
1. Navigate to Agent Creation
- Go to Dashboard → Workspaces
- Click on your workspace
- Click Agents tab
- Click "Create Agent" button
2. Basic Information
Agent Name (required)
- Clear, descriptive name
- Example: "Customer Support Bot", "Sales Assistant"
Slug (required)
- Auto-generated from name
- URL-friendly:
customer-support-bot - Used in URLs and API calls
Description (optional)
- Brief explanation of agent's purpose
- Example: "Answers customer questions about products and shipping"
3. System Instructions
This is the most important part - it defines your agent's personality and behavior.
What to include:
- Agent's role and purpose
- Tone and personality (friendly, professional, concise)
- What it can and cannot do
- How to handle specific situations
Example:
You are a helpful customer support agent for Acme Clinic.
Your role:
- Answer questions about appointments, services, and locations
- Be friendly, professional, and empathetic
- If you don't know something, offer to connect them with staff
Important:
- Never provide medical advice
- Always confirm appointment times carefully
- Be respectful of patient privacy
Tips:
- Be specific about what the agent should and shouldn't do
- Use examples if needed
- Keep it clear and concise (aim for 100-300 words)
4. Model Selection
Preferred Model Choose the AI model that powers your agent:
- gpt-4o-mini - Fast, cost-effective (recommended for most use cases)
- gpt-4o - More powerful, better for complex tasks
- claude-3-5-sonnet - Excellent for detailed, nuanced responses
- 120+ other models available
Model affects:
- Response quality
- Speed
- Cost per message
5. Advanced Settings (Optional)
Temperature (0.0 - 2.0)
- Lower (0.3) = Focused, consistent, deterministic
- Medium (0.7) = Balanced (default)
- Higher (1.5) = Creative, varied, less predictable
Max Tokens (response length)
- Default: 2048 tokens (~1500 words)
- Higher = longer responses (more expensive)
- Lower = shorter, concise responses
Chat Mode
- Streaming - Responses appear word-by-word (default, best UX)
- Non-streaming - Full response at once
- Structured - JSON output (for developers)
Conversation History
- Enabled - Agent remembers previous messages (default)
- Limit - How many past messages to include (1-20)
- More history = better context but higher cost
6. Rate Limiting (Optional)
Prevent abuse by limiting requests per user:
- Per minute (default: 10)
- Per hour (default: 100)
- Per day (default: 1000)
7. Review & Create
Review your configuration, then click "Create Agent".
Your agent is created instantly and ready to configure!
After Creation
Once created, you'll be redirected to the agent page where you can:
- Add Tools - Enable integrations and capabilities
- Upload Training - Add documents for knowledge
- Test in Playground - Try out your agent
- Deploy - Get widget code or share URL
Testing Your Agent
- Go to agent page → "Test in Playground"
- Chat with your agent and test:
- Does it understand its role?
- Is the tone appropriate?
- Does it handle edge cases?
- Refine system instructions if needed
Next Steps
Now that your agent is created:
- Train Your Agent - Upload documents and knowledge
- Add Tools & Integrations - Connect external services
- Deploy Your Agent - Share with users
Tips & Best Practices
✅ Do:
- Write clear, specific system instructions
- Test thoroughly before deployment
- Start with lower temperature for consistency
- Enable conversation history for better context
❌ Don't:
- Make instructions too vague or generic
- Choose expensive models unnecessarily
- Skip testing in playground
- Forget to set appropriate rate limits