HomeCourses › Build Your First AI Agent from Scratch

Build Your First AI Agent from Scratch

A weekend build: a terminal agent that plans, calls tools and remembers. No frameworks

1 phase. 6 lessons. 6 labs. Build a tool-calling AI agent from raw Anthropic API calls — no LangChain, no CrewAI. You write the tool layer, the planning loop, context management, and error handling yourself, then run it against 3 real tasks you actually chose.

6 lessons · 6 labs · 4 tools built · 1 real agent

Outcomes you will have by the end

What you will be able to do

Agent Architecture · Tool Schema Design · ReAct Planning Loops · Context Window Management · Retry & Error Handling

Every phase, every lesson, every project

The technologies you will use

Anthropic API · Python · httpx · Serper API

Roles this course prepares you for

Common questions

Do I need agent-framework experience?

No — this course is deliberately framework-free. You need Python (functions, classes, basic async/await) and an Anthropic API account. If you've used ChatGPT or Claude but never built a tool-calling agent, this is the right starting point.

Why no LangChain or CrewAI?

Frameworks are useful once you understand what they're wrapping. This course teaches the actual mechanism — the tool_use/tool_result loop — so that any framework you use later is legible to you instead of a black box.

How is this different from the Agentic AI Engineering course?

Agentic AI Engineering (c2) is a full 50-lesson, 10-phase production-focused track covering multi-agent systems, MCP, computer use, memory systems, and production infrastructure. This is a single 6-lesson weekend build that gets you one working agent, fast — a taste of the same ideas, not the deep track.

Will I need to pay for API usage?

Yes — the Anthropic API and Serper API (for web search) are both pay-per-use beyond a free tier, but a full run-through of this course costs a small fraction of a dollar in API usage.

What do I end up with?

A real Python agent on your own machine that calls 4 tools (web search, file reading, Python execution, URL fetching), plans multi-step tasks, manages context, and handles failures — run against 3 tasks you chose yourself, not toy examples.

Key terms in this course

Agent · Context Window · ReAct · Tool Use

Continue your learning path

Agentic AI Engineering

Start the Build Your First AI Agent from Scratch course

Create a free account — the opening phases of 24 of 30 courses are free, no credit card. Or see Pro pricing.

All courses · Pricing · About · FAQ · Glossary