Skip to main content
BlogResourcesPodcast
⚙️ Module 2 of 8 12 min read Phase 1: Foundation

Setup

Your Agent Workspace

Video Lesson Coming Soon

A video walkthrough for this module is in production. For now, dive into the written content below.

What You'll Learn

  • Choosing your AI model (GPT, Claude, Gemini)
  • Platform options and tradeoffs
  • The 5 workspace panels every platform shares
  • Basic configuration and cost awareness
In this module 7 sections

What You Need Before You Start

You do not need an expensive computer. You do not need a technical background. You do not need to know how to code. You need a computer, an internet connection, and about thirty minutes of setup time.

Here is what you are going to do in this module: choose your AI model, set up your agent workspace, and get to the point where you have an empty agent ready to train. This module just gets the infrastructure in place.

Step 1: Choose Your AI Model

Compare the leading AI models and their strengths.

🧠 Model Options
AI Models for Agent Development
💬
ChatGPT
Strong writing & reasoning, widely known, proven API
📖
Claude
Precise instruction-following, excellent long-context handling
🔍
Gemini
Research tasks, large information processing, multimodal
🔓
Open-Source
Free to run locally, lower cost, requires technical setup
0/4

Interactive — tap to explore

Your agent runs on an AI model — the engine that does the actual thinking and writing. Here are your main options as of early 2026:

ChatGPT (OpenAI): The most widely known. GPT-4o is the current general-purpose model. Strong at writing, reasoning, and following instructions. Available through the OpenAI API or through platforms that integrate it.
Claude (Anthropic): Known for following detailed instructions precisely and handling longer documents well. Strong at writing, analysis, and nuanced tasks. Available through the Anthropic API or through platforms that integrate it.
Gemini (Google): Strong at research tasks and processing large amounts of information. Available through Google AI Studio or the API.
Open-source models (Llama, Mistral, and others): Free to run but require more technical setup. Quality varies. Best for people with some technical comfort who want to avoid ongoing API costs.
💡
Budget Consideration

For most beginners, the practical choice is between ChatGPT and Claude, accessed through their APIs or through a platform that handles the integration. Both produce excellent results for freelance work when properly trained. Expect to spend somewhere between $20-$100 per month on model usage for a moderately active freelance agent.

Step 2: Choose Your Agent Platform

The AI model is the brain. The platform is the workspace where you configure your agent, manage its memory, connect its tools, and handle the flow of work.

Your options fall into three categories:

All-in-one agent platforms: These provide the workspace, model access, and tools in one package. They handle the technical integration so you can focus on training.
Build-your-own with open-source tools: OpenClaw is the most popular example — an open-source AI agent that runs on your computer and communicates through messaging apps. Powerful but requires comfort with technical setup. Its architecture maps directly to the Three Pillars: SOUL.md for personality, memory files for knowledge, TOOLS.md for capabilities.
Platform APIs with custom setup: Using the OpenAI or Anthropic API directly with custom code or no-code tools to build your agent workflow. Maximum flexibility, moderate technical requirement.
🧠
Getting Started

For most people taking this course, an all-in-one platform is the fastest path. You can always migrate to a more custom setup later once you understand the principles. Regardless of which platform you choose, the setup process follows the same general pattern.

Step 3: Set Up Your Workspace

Explore the essential building blocks of your workspace.

🌳 Workspace Structure
Core Components of Any Agent Platform
⚙️Agent WorkspaceSETUP
├─📝Instructions PanelPILLAR 1
└─💡System Prompt / Agent Instructions
├─🧠Memory/Knowledge PanelPILLAR 2
└─📚Reference Materials & Examples
├─🔧Tools PanelPILLAR 3
└─Skills & Integrations
├─🧪Testing & ChatTESTING
└─💬Direct Agent Interaction
├─🚀Deployment PanelCONFIG
└─🔗Automation & Connection Settings
0/11

Interactive — tap to explore

Every agent workspace has the same core components, even if they are labelled differently across platforms.

The Instructions Panel: This is where your system prompt goes — the RIDE framework document you will build in Module 4. Some platforms call this System Prompt, others call it Agent Instructions or Personality.
The Memory/Knowledge Panel: This is where your reference materials, examples, style guides, and accumulated knowledge live. Some platforms call this Knowledge Base, Context, Files, or Memory. This is Pillar 2.
The Tools Panel: This is where you enable or configure the capabilities your agent can use — web search, document reading, file creation, messaging. Some platforms call this Skills, Actions, Integrations, or Plugins. This is Pillar 3.
The Testing/Chat Interface: A way to interact with your agent directly to test it before going live. Every platform has some version of this.
The Deployment/Connection Panel: The settings that control how your agent receives and delivers work — whether through a marketplace, email, messaging platform, or other channel.
💡
Quick Reference

Spend ten minutes familiarising yourself with where each of these lives on your chosen platform. You do not need to understand every feature yet. You just need to know where the three pillars are configured.

Step 4: Configure the Basics

Before writing your system prompt, handle the basic configuration.

Name your agent: Choose something professional that reflects the service. Content Writer Pro tells a client more than My Agent 1. If the platform allows a description, write one sentence about what the agent does.
Select your AI model: If your platform offers model choices, start with the most capable model available to you. You can optimise for cost later once you understand the quality-cost tradeoff. Starting with the best model means your initial tests show you the ceiling of what is possible.
Set the automation level: If your platform supports it, set to semi-automated. This means the agent processes work but you review everything before delivery. Module 8 covers when and how to move to full automation.
Enable basic tools: Text generation (usually on by default) and communication (if available). Leave advanced tools disabled for now — you will add them only when needed.

Troubleshooting Common Setup Issues

💡
Troubleshooting

I cannot access the API: Most platforms require an API key from the AI model provider. You create this on the provider's website, usually under API Keys or Developer Settings. Copy the key and paste it into your agent platform's settings. If it does not work, check: is the key active? Does your account have billing set up? API access usually requires a payment method on file.

The interface looks different from what you described: AI platforms update frequently. The core components will be present — they may just be labelled or arranged differently. Look for synonyms: System Prompt equals Instructions equals Personality. Knowledge Base equals Memory equals Context equals Files.

I do not know which platform to choose: If you are paralysed by choice, pick any platform that offers a free trial and supports at least one of the major models (GPT-4o or Claude). The skills you learn transfer to any platform. You are not locked in.

I am worried about costs: Start small. Most platforms offer free tiers or trial credits. Your testing phase will give you a sense of how much each task costs. You can estimate monthly costs before committing to live work.

Your Foundation Is Ready

By now you should have an AI model selected, a platform set up, an empty agent created and named, configured with basic settings, connected to your chosen model, and set to semi-automated. You also have a rough sense of costs.

Ready to Go

Your agent is an empty shell. That is exactly right. Module 4 fills it with instructions. Module 5 fills it with memory. Module 6 tests it.

This module just made sure the shell is ready.

Key Takeaways

📝 My Notes
← The Landscape Three Pillars →