08. Agent Harness & Runtime
Build the runtime layer that turns an LLM into a reliable agent — loops, tools, permissions, MCP, and observability.
| | |
|---|---|
| Module ID | module-18 |
| Phase | Build |
| Lessons | 6 |
| Est. hours | ~5.5h |
Lessons¶
| # | Lesson | Duration | Level |
|---|---|---|---|
| 1 | What Is an Agent Harness? | 55 min | intermediate |
| 2 | Agent Loop and State | 60 min | advanced |
| 3 | Tools and Function Calling | 55 min | advanced |
| 4 | MCP — Model Context Protocol | 55 min | intermediate |
| 5 | Permissions and Safety in the Harness | 50 min | advanced |
| 6 | Observability in the Harness | 55 min | advanced |
Prerequisites: Course 07 · AI Agents — especially lessons 1–4 on the agent loop, ReAct, and tool use.
Start here: What Is an Agent Harness?