Getting started curriculum

ArcFlow workflows are written as simple declarations: you define agents, chain them into a workflow, and call run(). The Rust runtime handles scheduling, tools, memory, traces, and recovery. This curriculum teaches that model step by step: one concept per lesson, each with a copy-paste script you run locally on your machine.

Start with Install and build if you have not built the Python or TypeScript SDK yet.

Recommended order

Step	Track	Time (approx.)	Outcome
1	Fundamentals	30 min	Mental model: Agent, Workflow, `run()`
2	Writing agents	25 min	Instructions, roles, context policy
3	Writing workflows	35 min	Linear pipelines, graph intro, testing
4	Tools	25 min	Define tools, attach to agents, tool loop
5	Memory	20 min	Session, shared, vector setup
6	RAG	30 min	Ingest, retrieve, wire into agents
7	Integrating	25 min	Embedded SDK vs server, HITL, callbacks
8	Outcome paths	varies	End-to-end goals (static site, server API)

You do not need every track before shipping. After Fundamentals and Writing workflows, you can jump to an outcome path that matches your job.

Outcome paths (pick one goal)

Goal	Path
Fastest first run	First workflow in five minutes
Python app with optional live LLM	Python quickstart
TypeScript / Node service	TypeScript quickstart
HTTP integration (curl, backend team)	Server API quickstart
Public website chat widget	Static site chatbot

The ArcFlow pattern (every lesson uses this)

python

from arcflow import Agent, Workflow

agent = Agent(
 name="worker",
 role="assistant",
 instructions="Do one clear task with the user input.",
)

workflow = Workflow("my_flow")
workflow.step(agent)

result = workflow.run("your input here") # default: no API key required
print(result.output)

Layers you add later: more steps, Tool(...), MemoryConfig(...), provider=OpenAI(...), graph mode, server runtime=, HITL, external callbacks.

Where to go after this section

Need	Document
Reference depth	Guides
API signatures	Python SDK reference
Guided tracks A–H	Tutorials
Architecture	Concepts