Agents
AI agents are configured assistants that combine an LLM model with tools, data sources, and custom behavior.
What is an Agent?
An agent has:
- System prompt — defines personality and instructions
- LLM model — the model that powers the agent (or auto-routing)
- Tools — database queries, HTTP calls, MCP servers
- Triggers — webhooks that start conversations automatically
Agent Types
| Type | Use Case |
|---|---|
data_analyst | Query databases, generate reports |
customer_support | Answer questions, resolve issues |
sales | Lead qualification, product info |
general | General-purpose assistant |
Agent Lifecycle
- Draft — Being configured, not yet available
- Published — Active and accepting requests
- Archived — Disabled, preserved for reference
Managing Agents
Agents can be managed through:
- Console — Visual editor at console.zihin.ai
- MCP Server — AI-powered IDE integration (see MCP Server)
- REST API — Programmatic management (see Execution)