Writing agents
This track teaches how to define agents in Python and chain them into multi-step workflows. Each lesson is short, runnable with the built-in stub provider, and builds on the fundamentals series.
What you will learn
| Lesson | Topic |
|---|---|
| 01 Minimal agent | The three required fields: name, role, instructions |
| 02 Instructions that work | Writing prompts the runtime can act on |
| 03 Roles and multi-agent pipelines | How role frames behavior and how steps hand off output |
| 04 Context and prior steps | ContextPolicy basics: what each agent sees from earlier steps |
Prerequisites
Complete Install and build so from arcflow import Agent, Workflow works in your virtual environment. Read 03 Anatomy of a workflow first so terms like step, run input, and result.output are already familiar.
How these lessons are structured
Every page follows the same sections: Before you start, Concept, Example, Verify, and Next. Run each example as a standalone script. No API keys are required until you move on to Python quickstart.
After this track
| Goal | Next document |
|---|---|
| Verify trace events and status checks | Track A: First workflow |
| Full agent fields (tools, memory, provider) | Defining agents |
| Deep dive on context assembly | Context policies |