Skip to content
Trending storyDeveloping

Four categories of AI agent builders: workflow builders, frameworks, harnesses, runtimes

1 article1 sourcesince Oct 15Last article 3h ago ·

Overview

AISummary of 1 article

A post on X groups AI agent-building options into four categories: workflow builders, frameworks, harnesses, and runtimes.

The categories are sorted by how much of the agent the developer writes versus how much the platform runs. The author says vendors use the same terms for different things, giving LangChain's LangGraph as an example of a product it calls a runtime and Anthropic's Claude Agent SDK as one it calls a harness. The author adds that most production systems combine more than one approach.

Written by AI from the articles below · updated Oct 15, 9:34 PM ET

Check the sources:

Latest developmentOct 15, 8:04 PM ET
Four ways to build an AI agent, from workflow builders to runtimes

Article timeline

The articles in this story. Times are ET.

Oct 15
  1. Ishaan SehgalX
    Four ways to build an AI agent, from workflow builders to runtimes

    AIThe post groups AI agent-building options into four categories: workflow builders, frameworks, harnesses, and runtimes, based on how much of the agent the developer writes versus how much the platform runs. It says vendors use the same terms for different things, such as LangChain calling LangGraph a runtime and Anthropic calling its Claude Agent SDK a harness. The author notes that most production systems combine more than one approach.

Heat trend

Not enough continuous observations to show a trend yet.