Concepts

Start here if Harn is new to you, or if you've been writing Harn long enough that the docs you actually want are the ones that explain why things are shaped the way they are.

These pages don't teach you syntax and don't list every function. They explain the model that the rest of the docs assume you already have.

Mental model

The containment diagram for a Harn conversation: how llm_call, agent_loop, workflow, pipeline, and session fit together.

Glossary

Every term Harn uses for a conversational unit, with one-line definitions and pointers to the page that owns each one.

Choosing an abstraction

When to reach for llm_call, agent_loop, agent_turn, workflow_execute, spawn_agent, and friends.

Steering seams

Where you can safely inject a user message into a running agent, and where you can't (yet).

Coming from elsewhere

Terminology cross-reference for readers arriving from OpenAI Agents SDK, Anthropic Claude Agent SDK, LangGraph, Inngest, Mastra, ACP, A2A, and MCP.

If you're here to ship code today