07-23-2026

The Knowledge Work AI Stack Guide

A clickable guide to help those who are trying to level up their use of AI past chat make sense of how to map tasks to products. Use it in two directions.

Build a stack: click pieces and watch what can’t run together dim away — click and see that isn’t a harness you can run from there; click and see the environment drop out.

Start from a task: pick a task and a scenario to see a stack that supports it — “Hand it work today” with “Deep-research a market” prescribes just the surface and , work you can start with no setup at all.

Stacks for Tasks
Stack
Tasks

In order to take a high-level view of all of Anthropic’s Claude products, the widget leans on six groups of terms. It’s a lot to hold in one’s head — so here is every card defined. Click one for details and a “reach for it when”:

Surfaces

The glass you look through. A surface never does the work.

Environments

The place the agent loop executes. An agent can only reach what its environment can reach.

Harnesses

The software that runs the agent loop.

The connectors

What the agent can reach. Most are accounts you plug in; two come free.

The context

Every piece of text the agent reads. What you type is one piece; skills and memories are the rest.

The trigger

What starts the run.