Agentic engineering

Agentic engineering is the discipline around getting useful, trustworthy work from coding agents. It covers the system around the model: context, tools, specifications, harnesses, evaluation, review, and the judgment to know when automation should stop.

Start here

• Agentic Engineering — the shift from prompting to designing systems for agents.
• The Good Spec — specifications that give agents clear boundaries and verifiable outcomes.
• Agent Harness Engineering — why the surrounding runtime often matters more than the model.
• AI Evals — practical evaluation as an engineering discipline.

Build reliable loops

• Loop Engineering — designing feedback loops agents can safely run inside.
• Own the Outer Loop — keeping human judgment at the right level of abstraction.
• Agentic Code Review — reviewing machine-speed output without creating a human-speed bottleneck.
• Agentic Code Quality — preserving quality as autonomy increases.

Go deeper

• A practical course on AI agents
• The Agentic Engineering glossary
• Agent Skills
• Code Review in the Age of AI

Browse all topic guides