AI Agents
Learn to build agents that reason, use tools, and survive production reality.
Work from tool schemas and ReAct loops through MCP, orchestration frameworks, evaluation, and agentic coding platforms without treating the topic like hype-only content.
Focus
Tool-using agents
Move from chatbot-style prompting to systems that plan, call tools, and complete multi-step tasks.
Core topics
11 primary modules
Covers function calling, ReAct, frameworks, MCP, reasoning models, evaluation, and production patterns.
Outcome
Build one real agent
The section works best when you build and evaluate one end-to-end agent instead of skimming every framework.
Start here
Enter the phase intentionally
Use the intro and first modules to get the core mental model right before touching agent frameworks.
Open start here →Protocol and platform
Understand MCP and modern stacks
Jump to MCP, SDKs, reasoning models, and the 2026 platform landscape if you already know the basics.
Explore modern agent runtimes →Production bar
Measure and harden agents
Agent work becomes real when you evaluate trajectories, inspect tool use, and add safety and observability.
Go to evaluation →Phase map
Follow the agent curriculum in the right order
Foundations and tool use
Learn what agents are, how tool schemas work, and how reasoning plus acting changes the interaction model.
Frameworks, orchestration, and protocols
Compare agent frameworks, coordinate multi-agent workflows, and understand interoperability layers like MCP.
Reasoning, evaluation, and the current landscape
Study reasoning models, agentic platforms, evaluation methods, and coding-agent workflows.
Practice and assessment
Build one agent and measure it honestly
Assignment
Build a production-ready agent with tools, error handling, memory, and evaluation.
Challenges
Use the hands-on challenge set to pressure-test design choices and tool-use behavior.
Pre-Quiz
Assess baseline knowledge before you go deeper into frameworks and protocols.
Post-Quiz
Validate that you can reason about architecture, tool use, and production concerns.
Project ideas
- SQL agent: natural language to queries, results, and insights.
- Research agent: search, synthesize, and report with sources.
- Coding agent: requirements to code, tests, repair, and iteration.
- Support agent: retrieve context, respond, and escalate when needed.
What comes next
- Model evaluation and metrics
- Debugging and troubleshooting
- AI safety and red teaming
- AI-powered dev tools
Treat this phase as part of the production sequence in the repo: build one useful agent, inspect traces, evaluate trajectories, and then carry those lessons into safety and debugging.