-
Notifications
You must be signed in to change notification settings - Fork 0
Agents
Gen 3: This is a legacy Gen 2 article. For current GT AI OS 3.0 guidance, see gen3/agents.
Agents are customized AIs configured for specific tasks. This guide helps you find, use, and create agents in GT AI OS.
- Finding Agents - Browse and search the agent gallery
- Creating Agents - Build custom agents for your needs
- Configuring Agents - Edit and fine-tune agents
- Sharing Agents - Share with teams and export/import
An agent is a customized AI that combines:
- Personality: How it communicates (tone, style, approach)
- Knowledge: Documents it can reference from attached datasets
- Instructions: A system prompt that defines its behavior and expertise
For example, you might have a "Technical Support" agent that's formal and references your product documentation, and a "Brainstorming Partner" agent that's creative and conversational.
The Agents page has two views, toggled by the button in the top right corner:
| View | Purpose |
|---|---|
| Favorite Agents | Your landing page. Shows agents you've favorited for quick access. Click any agent to start chatting. |
| Agent Configuration | Browse all available agents, create new agents, and manage settings. |
- Click Agents in the sidebar
- Click any agent card to start a conversation
- The agent will respond based on its configuration
- Go to Agents and switch to Agent Configuration view
- Click Create Agent
- Fill in the name, description, and system prompt
- Save and start testing
See Creating Agents for a detailed guide.
Agents can be organized by category to help users find what they need. Categories are created by users - they're fully customizable to fit your organization's needs.
Add tags to agents for additional filtering. Tags help users find agents by keywords like "support", "technical", "sales", etc.
Star agents you use frequently to add them to your Favorites view for quick access.
- Start conversations from the Agents page - clicking an agent card is the fastest way to start chatting
- Use clear descriptions - help others understand what each agent does
- Test before sharing - verify your agent works well before making it available to others
- Attach relevant datasets - give agents the knowledge they need