Your First Agent¶

This guide walks you through creating, configuring, and testing your first AG2Trust agent.

Prerequisites¶

Before creating an agent, ensure you have:

• [x] An AG2Trust account
• [x] At least one LLM provider configured (Settings > Providers)

Creating an Agent¶

Via Dashboard¶

1. Navigate to Agents in the sidebar
2. Click the Create Agent button
3. Fill in the configuration form

Agent Configuration¶

Writing Effective System Prompts¶

Your system prompt defines how the agent behaves. A good system prompt includes:

1. Role definition - Who is the agent?
2. Capabilities - What can it help with?
3. Constraints - What should it avoid?
4. Tone - How should it communicate?

Example: Customer Support Agent¶

You are a customer support agent for TechCorp, a software company.

## Your Role
- Help customers with product questions
- Troubleshoot common technical issues
- Guide users through account management

## Guidelines
- Be friendly, professional, and empathetic
- Keep responses concise (2-3 paragraphs max)
- If you don't know the answer, say so honestly
- Never make up information about products or policies

## Escalation
If a customer asks about:
- Refunds or billing disputes
- Legal matters
- Account security concerns
Say: "I'll connect you with a specialist who can help with that."

Example: Code Assistant¶

You are a senior software engineer helping developers write better code.

## Expertise
- Python, JavaScript, TypeScript, Go
- Web frameworks (React, Vue, FastAPI, Express)
- Database design and SQL
- API design and best practices

## Response Format
- Explain your reasoning briefly
- Provide working code examples
- Include error handling
- Note any security considerations

## Style
- Be direct and technical
- Use code blocks with syntax highlighting
- Reference official documentation when helpful

Starting Your Agent¶

After creating an agent:

1. Find your agent in the Agents list
2. Click the Start button
3. Wait for status to change to Running (5-10 seconds)

Agent Status Lifecycle¶

stateDiagram-v2
    [*] --> Created: Create Agent
    Created --> Running: Start
    Running --> Idle: No activity (30s)
    Idle --> Running: New message
    Running --> Stopped: Stop
    Stopped --> Running: Start
    Running --> Error: Failure
    Error --> Running: Start (retry)

Testing Your Agent¶

Via Dashboard Chat¶

1. Click on your running agent
2. Use the built-in chat interface
3. Send test messages and verify responses

Via API¶

# Get your agent ID from the Dashboard
AGENT_ID="your-agent-id"
API_KEY="cust_your_api_key"

# Send a message
curl -X POST "https://agents.ag2trust.com/api/v1/agents/${AGENT_ID}/messages" \
  -H "X-API-Key: ${API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{"message": "Hello! What can you help me with?"}'

Expected Response¶

{
  "message_id": "msg_7f8g9h0i",
  "content": [
    {
      "type": "text",
      "text": "Hello! I'm here to help you with product questions, technical troubleshooting, and account management. What can I assist you with today?"
    }
  ],
  "metadata": {
    "tokens_used": 38,
    "model": "gpt-4o",
    "duration_ms": 892
  }
}

Agent Tools & Capabilities¶

AG2Trust agents can be equipped with various tools:

Available Tool Categories¶

Configuring Capabilities¶

Capabilities are predefined tool bundles:

code_review    - Git read-only + file read
file_operations - Full filesystem access
git_full       - All git operations
web_http       - HTTP client access
research       - Web search + HTTP

Configure capabilities when creating or editing your agent in the Dashboard.

Best Practices¶

1. Start Simple¶

Begin with a focused use case before adding complexity.

# Good: Focused
You are a code review assistant. Review code for bugs and suggest improvements.

# Too broad
You are an AI that can do anything the user asks.

2. Set Clear Boundaries¶

Define what the agent should and shouldn't do.

## Do
- Answer questions about our product
- Help troubleshoot common issues

## Don't
- Make promises about future features
- Share internal company information
- Provide legal or medical advice

3. Test Edge Cases¶

Before going live, test:

• [ ] Normal use cases
• [ ] Empty or very short inputs
• [ ] Very long inputs
• [ ] Off-topic questions
• [ ] Potentially harmful requests

4. Monitor Performance¶

Use the Dashboard to track:

• Response times
• Token usage
• Error rates
• User interactions

Troubleshooting¶

Agent won't start¶

1. Check provider status - Verify your LLM provider API key is valid
2. Review logs - Check agent logs in the Dashboard
3. Try a different model - Some models may have availability issues

Slow responses¶

1. Simplify the system prompt - Long prompts increase latency
2. Use a faster model - Consider gpt-4o-mini for speed
3. Check provider rate limits - You may be hitting API limits

Unexpected behavior¶

1. Review system prompt - Ensure instructions are clear
2. Check temperature setting - Lower for more consistent responses
3. Test in isolation - Rule out context from previous messages

Next Steps¶

• Create Agent Pools - Scale with load balancing
• Set Up Teams - Organize agent collaboration
• Configure Webhooks - Async response delivery
• API Reference - Full API documentation