- Databricks
- Free with an account
- Databricks Academy account
- Beginner
AI Agent Fundamentals
A free primer on agent concepts, useful before the MCP and agent courses.
Open course : AI Agent Fundamentals. Opens the provider site in a new tab
Engineers / Advanced
AI agents, MCP and evaluation
Agents with tools, the Model Context Protocol and the part most teams skip: an evaluation set and guardrails before production.
Who it is for
Engineers who have built an LLM prototype and now want an agent they can trust in production.
- Steps
- 9
- Level
- Advanced
What you can do afterwards
- Build an agent that calls tools through MCP
- Write an evaluation set and a regression gate for it
- Put a guardrail between the agent and a risky action
- Sit the Context Engineer Associate exam
Start here
Take the NexOps course
AI agents in production. Tools, memory, evaluation, guardrails, cost and operations, with the model of your choice
This path begins with our own course, with lessons, knowledge checks and labs. The external courses below are extra practice.
Extra practice
Course
Course
- Hugging Face
- Free
- Intermediate
AI Agents Course
A free agent course with smolagents, LlamaIndex and LangGraph, and a certificate on completion.
Open course : AI Agents Course. Opens the provider site in a new tab
Course
- Hugging Face
- Free
- Beginner
Model Context Protocol (MCP) Course
Learn the Model Context Protocol from architecture to a deployed MCP app, with certificates.
Open course : Model Context Protocol (MCP) Course. Opens the provider site in a new tab
Course
- LangChain Academy
- Free
- Intermediate
- 6 h
Introduction to LangGraph - Python
The official LangGraph course: state graphs, memory and human-in-the-loop, as an agent orchestration option next to Hugging Face.
Open course : Introduction to LangGraph - Python. Opens the provider site in a new tab
Course
- DeepLearning.AI
- Cost not stated
- Intermediate
- 15 h 7 min
Evaluating and Improving Agentic Applications
The evaluation course to pair with any agent work before it reaches production.
Open course : Evaluating and Improving Agentic Applications. Opens the provider site in a new tab
Build
An agent with one tool, one eval set and one guardrail
Goal. Build a small agent that uses one tool, and make it safe to change by testing it.
What to do
- Expose one useful read-only tool through an MCP server.
- Write 15 tasks with the result you expect, including three the agent should refuse.
- Run them on every change and fail the build when the pass rate drops.
- Add a guardrail that asks a human before any action that changes data.
What good looks like. You can change the prompt or the model and know within minutes whether the agent got better or worse.
Read
Posts to read
How we govern agents as a platform service, and how we start with agent memory in a governed pilot.
Check
AI harness check
Check whether your agent has test sets, regression gates and guardrails.
Certify
- Databricks
- Paid exam
- Intermediate
Databricks Certified Context Engineer Associate
The newest AI credential, on context engineering and governance for agent systems.
Course details come from each provider. Check the provider page for the latest.
Next path
- Engineers
- Advanced
Open-weight models and fine-tuning
Run and tune open models yourself: post-training, preference tuning, and when self-hosting is the right call.
View the pathLearn this with your team
We run workshops for company teams and for students, tailored to your stack.
Learn this with your team
Next step
Want your team to learn this together?
We give workshops to company teams and students, on-site or online. Tell us who the participants are and what they should be able to do afterwards.
Accepting projects for Q4 2026