Claude at Work: The Complete 2026 Field Guide to Getting Real Things Done and Built: Chat, Cowork, Code, Design, and Systems You Own

You have used the chat window. That is the smallest part of what Claude is now. In 2026, Claude is a family of models and working surfaces: a coworker that turns a folder of files into finished documents while you do something else; a builder that writes and tests real software from plain English; a design canvas that turns a description into a working prototype; agents inside Chrome, Excel, and Slack; and a platform layer of Skills, connectors, and APIs that assembles into systems you own rather than seats you rent. Claude at Work is the complete, honest, dated field guide to that whole toolkit, written for people who want real outcomes: owners, freelancers, operators, and professionals who are done dabbling and want finished work. Inside: which Claude to use for which job, and what each tier actually costs, in plain numbers with sources and dates. Chat done properly, past the prompt dump. Cowork, the agentic coworker, from its first job to governed keys. Claude Code for people who do not think of themselves as builders, anchored in the largest usage study Anthropic has published. Design without Figma, and where it honestly stops. The embedded surfaces, and when the standalone tools do better. Skills, MCP, and the API: the ownership layer that turns rented thinking into a loop you keep. Every product carries its release status in plain words. Every number prints its source, its date, and its denominator. Nothing is oversold, because the book's usefulness is the sales pitch. It was also written by the thing it is about: drafted by Claude inside a gated pipeline, checked by a human against sources the machine did not write, and signed by a human at the end. The colophon shows the receipts. Rent the thinking. Own the loop. A human still signs. From the author of A Human Still Signs and Your Hermes Agent.