← NALO SEED
AI ECOSYSTEM · INTERACTIVE TRAINING TOOL
A visual companion to AI Foundations

The Anatomy
of an Agent.

An AI agent is not a single thing. It is a layered system. Click any part of the diagram below to see what each layer does and why it matters.

THE HARNESS CRM DRIVE STRIPE GMAIL MODEL THE BRAIN MCP · EXTERNAL SYSTEMS CLICK ANY ELEMENT
↑ Each element is interactive · Click to explore
// 02 / DEMO

Watch the system work.

A request from a user flows through every layer. Press play to see how the model, harness, tools, and skills work together to answer one question.

User "Send Jane the proposal we discussed yesterday."
01
Harness
Receives the request, pulls in relevant context (yesterday's conversation, who Jane is) and hands it to the model.
02
Model
Reads the context and decides: this is a proposal-sending task. The right skill applies.
03
Skill
"Send Proposal" skill loads: find the doc, verify the recipient, draft the email, send it.
04
Tools
Model calls search_drive("proposal Jane") through MCP. Drive returns the file.
05
Tools
Model calls lookup_contact("Jane"). CRM returns her current email.
06
Model
Drafts the email body using yesterday's context and the proposal content.
07
Harness
Guardrail check: this is an external send. Pause and ask the user to confirm before calling send_email.
08
Tools
User confirms. Model calls send_email. Done in 14 seconds.
// 03 / WITH AND WITHOUT

Why every layer matters.

Take any one of the five layers away and the system breaks in a specific way. This is a quick way to explain to a client what they are paying for.

01

Model

Without Nothing. No language understanding, no reasoning, no output. There is no system.
With Can read, reason, write, and translate. Cannot do anything else by itself.
02

Harness

Without A chat window. Forgets every conversation. Cannot take any action. Cannot recover from errors.
With A worker that remembers, runs long tasks, handles errors, and orchestrates everything.
03

Tools

Without All talk. The agent can describe what to do but cannot do anything in the real world.
With Hands. Sends email, updates records, runs reports, books appointments, takes payment.
04

Skills

Without Drift. The agent improvises every time. Quality varies. Same task takes 10x longer.
With Consistency. Repeatable tasks run the same way every time, at the same quality.
05

MCP

Without Bespoke integrations. Every new tool is a custom build. Fragile. Expensive to maintain.
With Plug and play. Connect to thousands of systems through a single standard. Future proof.
// 04 / GLOSSARY

Vocabulary that pays off.

The terms that come up most often in real conversations with vendors and clients. Worth knowing cold.

Agent
An AI system that can take actions in the world. Not just generate text but call tools, make decisions, and follow up.
Agent Loop
The core runtime cycle. The model thinks, calls a tool, reads the result, thinks again, repeats until done.
Context Window
How much information the model can hold in its head in a single call. Finite. Filling it with junk degrades quality.
Context Rot
The slow degradation of agent performance as a session gets longer and accumulates irrelevant information.
Evals
Tests that measure whether an agent actually works. Without them you cannot tell if changes made things better or worse.
Hallucination
When a model confidently makes up information. Mitigated by giving it access to real data through tools.
Harness
The software wrapper that turns a model into an agent. Includes the loop, tools, memory, and context management.
MCP
Model Context Protocol. The open standard for connecting AI agents to external systems. Anthropic created it; the industry adopted it.
Memory
Persistent storage that lets an agent remember things between sessions. Lives outside the context window.
Model
The large language model itself. Claude, GPT, Gemini. The brain of the system.
RAG
Retrieval-Augmented Generation. The pattern of pulling relevant documents into the prompt before the model answers.
Skill
A reusable instruction pack that bundles a workflow plus its supporting files. Loaded when relevant.
Subagent
A specialized agent spawned by a parent agent to handle a specific subtask in isolation.
Token
The unit a model reads and writes. Roughly three-quarters of a word in English. Used for billing.
Tool
A function the model can call to take action. The agent's hands.
Tool Call
A structured request from the model asking the harness to run a specific function with specific inputs.