1.6 Foundations
Definition
- AI Agents by Berkley Book. N8N Blog on AI Agent Flows.

flowchart TD
%% --- Styles (safe) ---
classDef env fill:#f7e8f0,stroke:#c2185b,stroke-width:2px,rx:12,ry:12,color:#111;
classDef agent fill:#e8f7f0,stroke:#0f9d58,stroke-width:2px,rx:12,ry:12,color:#111;
classDef comp fill:#e6f0fb,stroke:#1a73e8,stroke-width:1.5px,rx:8,ry:8,color:#111;
classDef action fill:#fff5cc,stroke:#eab308,stroke-width:1.5px,rx:8,ry:8,color:#111;
%% --- Environment ---
subgraph ENV["Environment"]
PERCEPTS["Percepts"]
ACTIONS["Actions"]
end
class ENV env
%% --- Agent ---
subgraph AG["AI Agent"]
S["Sensors<br/>(inputs, APIs, webhooks)"]
PERCEPTION["Perception Layer<br/>(preprocess, embed)"]
KB["Knowledge Base / Memory<br/>(facts, vectors, state)"]
REASON["Reasoning Engine<br/>(LLM+prompts, rules, RL)"]
PLAN["Planning & Goals<br/>(decompose, schedule)"]
LEARN["Learning Module<br/>(feedback, fine-tune, RL)"]
ACT["Actuators<br/>(APIs, commands, outputs)"]
end
class AG agent
class S,PERCEPTION,KB,REASON,PLAN,LEARN,ACT comp
%% --- Flow ---
PERCEPTS --> S --> PERCEPTION --> KB
PERCEPTION --> REASON
KB --> REASON
REASON --> PLAN --> ACT
LEARN --> KB
LEARN --> REASON
ACT --> ACTIONS
- Definition by Harrison, LangChain Founder
- Ambient Agents: Ambient agents listen to an event stream and act on it accordingly, potentially acting on multiple events at a time)
- 12-Factor Agents - Principles for building reliable LLM applications
Agentic Architecture
- Agents Intro - a Google Whitepaper
- Building Effective Agents - Anthropic
- Cohere - How enterprises can start building Agentic AI
- AI Agents vs. Agentic AI White paper


