powers
LLM
Model capabilitiesModel limitationsModel selectionStructured outputs
Agentic AI Learning Map
Learn the terms behind agentic AI development.
A developer-facing map for understanding LLM agents: definitions, relationships, learning priority, and the first practical step to take. Use the term lookup when you want a direct "Should I care?" answer.
Agent-centered system map
Agentic AI is not just an LLM. An agent is a system around an LLM: context grounds it, tools let it act, a control loop drives it, and guardrails keep it safe.
grounds
Context
InstructionsUser/task context
drives
Control Loop
PlanActObserveDecideStop / retry
lets act
Tools
APIs
constrains
Guardrails
Safety checks
Selected term
Agent
Definition
An agent is an AI system that combines an LLM with context, tools, memory or state, control logic, and guardrails to pursue a task over one or more steps.
Role in the Agent model
Agent is the central system: the LLM powers it, context grounds it, the control loop drives it, tools let it act, and guardrails constrain it.
Common confusions
- An agent is not just an LLM; it is the system around the LLM.
- Autonomy is a spectrum; a useful agent may still have strict permissions and human approvals.
- Tool use alone does not guarantee good agent behavior without context, state, control logic, and guardrails.
Key subtopics
Connected concepts
5 key links- uses
powers it
- enables
grounds it
- uses
drives it
- enables
let it act
- protects
constrain it
11 more related concepts are kept out of this summary to keep the report readable.
Term Directory
Search by term, alias, category, or related technology.