Agent Concepts
Foundational concepts for understanding how AI agents work.
Understanding these core concepts will help you design and build effective AI agents.
Core Concepts
What Makes an Agent "Agentic"
An AI agent differs from a simple chatbot in its ability to:
- Reason about tasks and break them into steps
- Decide which tools or actions to take
- Execute actions through tool calls
- Observe results and adjust behavior
- Iterate until the task is complete
This iterative loop is what makes agents autonomous and capable of handling complex, multi-step tasks.
Key Components
Every AI agent in Axellero consists of:
| Component | Purpose | Example |
|---|---|---|
| LLM | Reasoning and language understanding | GPT-4, Claude, etc. |
| System Prompt | Defines agent behavior and constraints | "You are a helpful assistant..." |
| Tools | Actions the agent can perform | Search, calculate, query database |
| Memory | Context retention across turns | Chat history, tool results |
| Knowledge Base | Document retrieval (RAG) | Product docs, FAQs |
Agent vs Workflow
| Aspect | AI Agent | Workflow |
|---|---|---|
| Decision Making | Dynamic (LLM decides) | Static (predefined paths) |
| Execution Path | Non-deterministic | Deterministic |
| Best For | Conversational, exploratory tasks | Data pipelines, scheduled jobs |
| Tool Usage | Decided at runtime | Defined at design time |
| Cost | Variable (depends on reasoning) | Predictable |
Use AI Agents when:
- Tasks require natural language understanding
- The execution path depends on user input
- Dynamic tool selection is needed
Use Workflows when:
- Steps are known in advance
- Deterministic execution is required
- Cost predictability is important
Learn More
Agentic Loop
How agents reason and act iteratively
Multi-Agent
Orchestrating multiple agents together
Guardrails
Content moderation and policy enforcement
Implementation Guides
Ready to build? Move to practical implementation:
- Tools - Connect JavaScript, HTTP, database, and MCP tools
- Knowledge Base Node - Add document context with RAG
Technical Reference
For advanced configurations and node specifications:
- AI Agent Node - Technical parameters and loop control
- AI Sub-Agent Node - Multi-agent orchestration patterns
- Knowledge Base Node - Advanced RAG optimization