Create Agent with Agent¶
Let Agents design and create other Agents through natural conversation.
Overview¶
Instead of manually configuring Agents, use the Agent Creator built-in tool to have Agents help you design specialized Agents through guided conversation.
| Capability | Description |
|---|---|
| Conversational Design | Agent asks questions to understand your needs |
| Smart Recommendations | Suggests models, tools, and configurations |
| Learn from Examples | Export existing Agents to study successful patterns |
| Iterative Refinement | Adjust designs based on your feedback |
How It Works¶
graph LR
User[You describe needs] --> Agent[Agent Creator]
Agent --> Questions[Asks clarifying questions]
Questions --> Design[Designs configuration]
Design --> Review[You review & feedback]
Review --> Refine{Satisfied?}
Refine -->|No| Design
Refine -->|Yes| Register[Agent registered]
Typical conversation:
- You: "I need an Agent to review Python code"
- Agent asks about: code style focus, security checks, which model to use
- Agent designs configuration and creates draft file
- You review and suggest changes
- Agent refines until you approve
- Agent registers new code reviewer to system
Why Use Agent Creator¶
| Without Agent Creator | With Agent Creator |
|---|---|
| Learn technical configuration format | Natural conversation |
| Look up parameter meanings | Agent explains options |
| Guess appropriate settings | Agent recommends based on needs |
| Start from scratch | Learn from existing Agents |
Quick Example¶
You: "Create an Agent for data analysis"
Agent Creator:
- Asks about data types, analysis methods, output format
- Gets design specification
- Suggests models and tools (file tools, memory tools)
- Creates agent/data_analyst.json configuration
- Waits for your approval
- Registers Agent when you're satisfied
Related Documentation¶
Full details: Agent Creator Tool
Related: - Agent Configuration - Understand Agent parameters - Built-in Tools - All available system tools