What you'll be able to do

✓Run a tool-using agent loop and read its thought, action and observation trace
✓Describe tools to a model as a schema and send them in the message format hosted models use
✓Compare how two agent development kits wrap the same loop
✓List and call a tool through a real MCP server
✓Reproduce four agent failures and check the guard that stops each

See what an agent actually is, a short loop around a model, and the few guards that decide whether it is safe to give it tools.

⌁ The same loop sits inside tool-using assistants; agent development kits and the Model Context Protocol are common ways to connect a model to tools.

Start this internship

Create an account to unlock the 4 sections, the workbench, and askThili.

1. Why this course: from the agents in your syllabus to the agents being built now

🔒 locked

2. The Loop, by Hand

🔒 locked

3. The Same Loop with a Real Model, and the Kits

🔒 locked

4. Make It Fail, Then Guard It

🔒 locked

Agents: A First Look

Sections