Phase 10 of 12 · about 10 hours of study
Agent harnesses
Build the software around a model that turns it into a capable, safe agent. 14 lessons, about 10 hours of study, in four tracks. One lesson is free to try without an account: the first, What an agent harness is. The rest need a free account.
Start with What an agent harness is Create a free account
What you'll learn
Harness foundations
- What an agent harness is FreeThe software around a model that turns it into an agent.
- The core loop, built properlyStop reasons, parallel tool calls, errors and turn budgets.
Tools & environment
- A tool registry for agentsSchemas from functions, validation, output limits and helpful errors.
- File and search toolsRead, write, edit by exact replacement, glob and grep inside one folder.
- Running code safelyExecution tools, step limits, output caps and real sandboxes.
Context & control
- Permissions and approvalsRules, modes and approval prompts that keep a person in control.
- Hooks and skillsExtend a harness at lifecycle events, and load know-how on demand.
- Managing the context windowBudgets, clearing old tool results, compaction and a cacheable prefix.
- Memory, plans and long tasksTo-do lists, progress notes, memory files and checkpoints.
Scaling & shipping
- SubagentsHand a focused task to a fresh context with fewer tools.
- Streaming, interrupts and sessionsLive output, stopping safely, and saving and resuming transcripts.
- Evaluation harnessesRun agents on tasks in fresh environments and grade what they leave behind.
- Tracing, budgets and guardrailsSee what an agent did, cap what it spends, and stop it when it loops.
- Capstone: build a coding agentA harness that edits a project, runs its tests and stops when they pass.
Before you start
Lessons here build on earlier phases:
- Phase 1: Building with LLMs
- Phase 6: Modern LLM engineering
- Phase 7: Software engineering
- Phase 9: Model Context Protocol
The placement test can mark off what you already know, so you only study what's new to you.
What you'll build
A harness that edits a project, runs its tests and stops when they pass. Read the tutorial.
- Combine the loop, tools, permissions, a stop check and an event log into one coding agent
- Give the agent file tools that stay inside the project and refuse to change the tests
- Make the harness, not the model, decide when the work is done: when the tests pass
When you finish
Finish every lesson in this phase, or show you know it in the placement test, and you earn a certificate of completion, "Agent harnesses (Phase 10)", with its own link you can share.