Learn AI agents by doing the work.
Start with one bounded workflow. Then choose the Operator path to use existing agents safely or the Builder path to create agentic systems of your own.
The route
Shared foundation
Understand · Design · Bound
Operator
Use agents for real work with review and control.
Builder
Build, connect, evaluate, and operate agent systems.
How it works
Every lesson ends with evidence.
Learn
Understand one durable idea without depending on a product interface.
Build
Complete a small task with explicit inputs, actions, and expected results.
Verify
Use a checklist, source, diff, or test result to decide whether it worked.
Course roadmap
Build the foundation, then choose your path.
The paths are not exclusive. Many people become confident Operators first, then return to the Builder path when a real workflow needs custom software.
Shared foundation
What Is an AI Agent?
Learn the practical difference between a chatbot, a fixed workflow, and an agent that can choose tools and adapt its next step.
Design Your First Agent Workflow
Turn one recurring task into a small, testable agent workflow before adding integrations or automation.
Operator path
Use existing agents to complete real work safely, repeatably, and with clear review points.
Use a Coding Agent on a Real Project
Use a coding agent to inspect, change, and verify a small project while keeping every modification reviewable and reversible.
Choose Your First AI Agent Stack
Choose a minimal model, interface, tool, data, and review stack based on one workflow instead of collecting disconnected AI products.
Connect tools with MCP
Add data and actions without turning every integration into custom glue code.
Build a repeatable operating playbook
Turn successful prompts, approvals, and checks into a reusable team workflow.
Builder path
Create agentic systems with explicit tools, context, evaluation, permissions, and recovery.
Agent Safety: Permissions, Evidence, and Recovery
Design the permission boundaries, approval checkpoints, evidence trail, and recovery plan every useful agent workflow needs.
Tool calling and structured output
Give an application reliable actions and machine-readable results.
RAG, memory, and context
Choose what the agent should retrieve, retain, and forget.
Single-agent and multi-agent patterns
Add orchestration only when one bounded agent is no longer enough.
Evaluation and production operations
Measure accepted outcomes, cost, latency, failures, and recovery.
Start small
One useful workflow beats ten agent demos.
Begin with a reversible task, require evidence, and add autonomy only after the workflow works under review.