3.5 Module 3 · Single-Agent Design Patterns

Build Your First Agent

Guided step-by-step walkthrough to construct a research agent from scratch. Choose your platform, define tools, write instructions, and run your first successful query.

Guided Agent Builder First-Run Diagnostic ChatGPT Claude

Guided Agent Builder

Follow each step to configure your research agent. Your choices determine the system prompt, tools, and behaviour of the agent you will test in the diagnostic below.

1
2
3
4
Step 1 of 4

Choose Your Platform

Each platform has different strengths for building agents. Select one to see platform-specific instructions throughout the builder.

OpenAI
GPT-4.1 + Assistants API. Best for structured tool calling and JSON mode.
Anthropic
Claude + Tool Use. Best for long context, nuanced reasoning, and safety.
Zapier
No-code agent builder. Best for connecting apps without writing code.

First-Run Diagnostic

Watch your agent process a real query. The diagnostic shows the full reasoning trace, tool calls, intermediate results, and final output.

Complete the Guided Agent Builder above to unlock the diagnostic.

Key insight: A well-built agent is not about the most powerful model. It is about clear instructions, well-defined tools, and a specific scope. A focused agent with 3 tools will outperform a vague agent with 20 tools every time.