Skip to content

Agents Creating

GT AI OS Release edited this page Jun 18, 2026 · 3 revisions

Creating Agents

This guide walks you through creating custom agents in GT AI OS step by step.

Prerequisites

  • You need appropriate permissions to create agents
  • Understand what task the agent should accomplish
  • Have a clear idea of the agent's personality and behavior

Creating a New Agent

Step 1: Open the Creation Form

  1. Go to Agents in the sidebar
  2. Click the toggle to switch to Agent Configuration view (if in Favorites view)
  3. Click Create Agent
  4. The agent creation form opens

Step 2: Basic Information

Fill in the required fields:

Field Required Description Example
Name Yes Clear, descriptive name "Customer Support - Technical"
Description Yes What this agent does "Helps users troubleshoot technical issues"
Category Yes Select or create a category "Technical", "General", "Creative"
Tags No Keywords for filtering "support", "technical", "help"

Note: Categories are customizable - you can create new ones or use existing categories your organization has set up.

Step 3: System Prompt

The system prompt is the most important part - it defines your agent's behavior.

Structure your prompt like this:

# Role
You are a helpful customer service agent for [Company].

# Responsibilities
- Answer questions about products and services
- Help resolve common technical issues
- Escalate complex problems to human support

# Communication Style
- Be professional and empathetic
- Use clear, simple language
- Provide step-by-step instructions when helpful

# Boundaries
- Don't make up information you don't know
- Don't discuss competitors negatively
- Redirect off-topic questions politely

Best practices for system prompts:

  • Be specific about the agent's role and expertise
  • Define the tone (formal, friendly, technical)
  • List clear responsibilities
  • Set boundaries on what the agent should and shouldn't do
  • Include examples if helpful

Step 4: Model Configuration

Configure how the AI generates responses:

Setting Range Description
Temperature 0.0-2.0 Controls creativity vs. consistency

Temperature guide:

  • 0.0-0.3: Consistent, factual responses (technical support, Q&A)
  • 0.4-0.7: Balanced (general use, default is 0.7)
  • 0.8-2.0: Creative, varied responses (brainstorming, writing)

Step 5: Attach Datasets

Give your agent specialized knowledge:

  1. In the Datasets section, use the search box to find datasets
  2. Check the boxes next to datasets you want to attach
  3. Use Select All or Clear All for bulk selection
  4. Selected datasets appear as badges above the list

Tips:

  • More datasets isn't always better - keep it focused
  • Update dataset attachments when your documents change
  • Test retrieval quality after attaching

Step 6: Easy Buttons (Optional)

Add quick-access conversation starters that appear as buttons in the chat:

  1. Find the Easy Buttons section
  2. Click Add Easy Button to create a new button
  3. Enter the prompt text that will be sent when clicked
  4. Add up to 10 easy buttons

Examples:

  • "Summarize this document"
  • "Explain this in simple terms"
  • "Help me draft an email about..."

Step 7: Visibility Settings

Control who can access your agent:

Visibility Who Can Access Best For
Individual Only you Personal agents, development
Team Selected team members Team-specific workflows
Organization All users in your organization Company-wide tools

Note: Only Tenant Admins can set Organization visibility. Once set, all users can access the agent. See Sharing Agents for details.

When selecting Team visibility:

  1. Choose which teams should have access
  2. Each team's members will see the agent

Step 8: Save and Test

  1. Review all your settings
  2. Click Create Agent
  3. The agent is created and available immediately
  4. Click on the agent to start a test conversation
  5. Iterate based on results

Testing Your Agent

Initial Testing

After creating:

  1. Try the main use cases you designed for
  2. Test edge cases and unexpected inputs
  3. Verify the tone matches your intentions
  4. Check that datasets are being used correctly

Iteration Cycle

  1. Note any issues in responses
  2. Edit the agent to adjust
  3. Test again
  4. Repeat until satisfied

Tips for Better Agents

Write Clear Instructions

Be explicit about:

  • What the agent should do
  • How it should communicate
  • What it should avoid
  • How to handle edge cases

Start Simple

  • Begin with basic functionality
  • Test thoroughly before adding complexity
  • Iterate based on actual usage

Use Datasets Wisely

  • Attach only relevant datasets
  • Keep dataset content focused
  • Update datasets when information changes

Test Edge Cases

  • Try unexpected inputs
  • Test boundary conditions
  • Verify error handling

Common Mistakes to Avoid

Mistake Problem Solution
Vague system prompt Inconsistent responses Be specific about role and behavior
Too many datasets Slow, unfocused retrieval Only attach what's needed
Wrong temperature Responses too rigid or chaotic Match setting to use case
No testing Sharing broken agents Always test before sharing
No boundaries Off-topic responses Define what agent shouldn't do

Next Steps

After creating your agent:

Clone this wiki locally