Learn to build
AI agents.
From scratch. For real.
Hands-on courses on orchestration, tool use, MCP, and production agent systems. Bring your own model. No frameworks until you understand what they abstract.
4
tracks
20
modules
67
lessons
100%
free
# Your first agent - built from scratch, no frameworks
import ollama
def run_agent(goal, tools, registry, max_steps=10):
messages = [{"role": "system", "content": f"Goal: {goal}"}]
for _ in range(max_steps):
response = ollama.chat(
model="llama3",
messages=messages,
tools=tools,
)
message = response.message
messages.append(message)
# No tool calls means the model is done.
if not message.tool_calls:
return message.content
# Otherwise run each tool and feed the result back.
for call in message.tool_calls:
result = registry[call.function.name](**call.function.arguments)
messages.append({
"role": "tool",
"content": str(result),
"tool_call_id": call.id,
})Curriculum
Four tracks to mastery
Each track builds on the previous. Start with fundamentals and work your way to deploying production agent systems.
Agent fundamentals
The orchestration loop, tool use, ReAct, memory, context engineering, and agentic RAG.
Orchestration patterns
Multi-agent systems, topologies, state machines, safety controls, and metacognition.
Building with MCP
Protocol fundamentals, building servers, building clients, and production MCP.
Production agents
Reliability, evaluation, observability, deployment, and scaling.
What makes this different
Built by an engineer shipping agents at scale
Not another LangChain tutorial. Learn the concepts, then choose your tools.
Build from scratch
No frameworks until you understand what they abstract. Every concept implemented in plain Python first.
Interactive sandboxes
Write and run Python directly in your browser. Pyodide handles the logic exercises with no local setup required.
No vendor lock-in
Provider-agnostic. The patterns work with any LLM API: Anthropic, OpenAI, a local model, or whatever ships next.
Production-tested patterns
Orchestration loops, MCP servers, observability dashboards. Patterns used in real systems with 200+ engineers.
Progressive depth
Start with a 30-line agent, end with a production multi-agent system. Each track builds naturally on the last.
Zero cost to learn
Free platform, free exercises, free curriculum. Bring whichever model you already have access to.
Prerequisites
You'll need three things to get started.
Python basics
Functions, classes, and comfortable with pip.
Command line
Navigate directories, run scripts, install packages.
An LLM you can call
Anthropic, OpenAI, local Ollama, anything that takes a chat completion. Optional for early lessons.
Ready to build your first agent?
Start with Track 1 and have a working agent in 30 minutes. No sign-up required to browse the curriculum.