Learn by topic
Guides
Each guide teaches one idea in depth, with its own examples. Read the one you need before the problem that uses it. They explain how things work, and they do not give away any problem's answer. New here? Start with Start here, then read these in order.
- 01How an agent loop works7 min read
What a model can actually see, how a harness feeds it, and why a small loop sits at the heart of every agent.
- 02Tools and schemas7 min read
How a model uses tools, why the description you give it is the only thing it knows, and how to make results it can use.
- 03Context engineering6 min read
The model treats everything in the conversation as evidence. Learn how to decide what goes in, what stays out, and why a stronger prompt cannot undo a bad line.
- 04Evaluating agents7 min read
Why checking the final answer is not enough, how to judge the way a run got there, and how to write a grader you can trust on runs you have not seen.
- 05Reliable tool use6 min read
Tools fail loudly, fail quietly, and sometimes return confident nonsense. Learn how to tell the difference and build a loop that does not believe everything it reads.
- 06Designing a tool set6 min read
More tools is not more capability. Learn how to choose a small menu, name and describe each tool so the model picks the right one, and hide the mess behind wrappers.
- 07Agent memory6 min read
A new session starts with an empty context. Learn what memory is, what deserves saving, how to name it so it can be found, and how to read it back without dumping everything.