Agents
Agents are AI-powered chat assistants that can answer questions, perform tasks, and use tools on your behalf. Each agent is configured with a personality, capabilities, and access to specific tools and knowledge bases.
Creating an Agent
Click the + Create button to build a new agent. You'll configure:
- System prompt — defines the agent's personality, behavior, and instructions
- LLM model — select from available models (GPT-4, Claude, etc.)
- Tools — attach tools the agent can call to perform actions
- Knowledge bases — connect documents the agent can search and reference
- Data sources — link external data the agent can access
Agent Components
| Component | Purpose |
|---|---|
| System Prompt | Defines how the agent behaves, its tone, and what it should or shouldn't do |
| Tools | Functions the agent can invoke — API calls, database queries, calculations |
| Knowledge Bases | Document collections the agent searches for context (RAG) |
| Data Sources | External systems the agent can read from via MCP connectors |
| Skills | Reusable instruction patterns that enhance the agent's capabilities |
Managing Agents
From the agent card menu (three dots), you can:
- Assign to role — make the agent available to users with a specific role
- Assign to user — share the agent with a specific user
- Publish — make the agent available organization-wide
- Export as template — save the agent configuration as a reusable template
- Change organization — move the agent to a different organization
- Delete — permanently remove the agent
Quick Tips
- Write clear, specific system prompts — they directly affect response quality
- Attach only the tools your agent actually needs to reduce confusion
- Test your agent in the chat interface before publishing to your organization
- Use templates to quickly create agents from proven configurations
- Add knowledge bases for domain-specific Q&A
- Use versioned data sources to keep your agent's context up to date