Course 07
Agents & MCP
The agent loop as a mechanism you could draw, and everything wrapped around it: tools, MCP, context, permissions, evaluation and the bill.
Written for someone whose agent already works on the happy path and falls over everywhere else. You can build one; you cannot yet say why turn three called the wrong tool, where the tokens went, or what happens when the page it reads tells it to do something.
- 13
- modules
- 60
- lessons
- 145
- curated videos
- 51h 19m
- of video
What an agent actually is
Replace the word 'agent' with a mechanism you could draw on a whiteboard.
Tools and function calling
Give a model hands, and make the hands safe: function calling, tool schemas, validating arguments, errors and retries, and having too many.
MCP: the protocol
Understand Model Context Protocol as a wire format, not a product.
Building an MCP server
Ship a server of your own and know why each part is there: a first server, exposing resources, testing it, and packaging it for other people.
Context engineering
The loop's real constraint is the window. Manage it deliberately.
Planning and decomposition
When a plan beats a bigger prompt: plan-then-execute, scratchpads, subtasks and checkpoints, and the cases where planning makes an agent worse.
Multi-agent orchestration
More agents is a design choice with a bill: when more than one is warranted, handoffs, subagents, orchestrator-worker, and committee cost.
Permissions and sandboxing
Decide what the loop may touch before it touches it: least privilege, sandboxed execution, human-in-the-loop gates, and handling credentials.
Untrusted input in a loop
The security model that changes once the model can act: prompt injection through tools, the lethal trifecta, OWASP's LLM Top 10, and real defences.
Evaluating agents
Judging a process, not an answer: why output eval falls short, building a test set, trajectory eval, LLM-as-judge, and regression before deploy.
Cost, latency and caching
Where an agent loop burns money and how to stop it: where tokens go, prompt caching, model routing, latency in a loop, and circuit breakers.
Observability and debugging runs
Reading a trace of something that will not repeat: what to log, tracing a run end to end, reproducing a bad run, and monitoring one in production.
Shipping an agent
Put it in front of people and keep it there: scoping what it may do, building a real one, the failure modes users see, and rolling out safely.
Curated from 15 channels
Every video is a public YouTube video. We pick the single clearest explanation for each lesson and credit the channel that made it.