Claude moves beyond chat

Anthropic is no longer treating Claude like a chatbot that waits for a prompt and replies with text. Its current product stack now includes browser control in Chrome, code execution, managed agents in the application programming interface, and organization-level admin tools for companies running Claude at work. (claude.com, platform.claude.com, platform.claude.com) The clearest sign is the new Chrome extension. Claude in Chrome can read pages, click buttons, fill forms, and navigate websites from a side panel, and Anthropic says it is in beta on paid plans including Pro, Max, Team, and Enterprise. (support.claude.com, claude.com) Anthropic is tying that browser tool directly to coding work. Claude Code can now connect to Chrome so a developer can build in the terminal, open the app in a browser, inspect console errors and page structure, and then send fixes back into the code without switching tools. (code.claude.com, support.claude.com) That changes what “assistant” means. Instead of answering a question about a bug, Claude can test a live web app, read the failure, and keep iterating while sharing the browser session you are already logged into, pausing only for things like a login page or a CAPTCHA test. (code.claude.com) Anthropic is also building the plumbing underneath that behavior. Its developer docs now split Claude into direct model access through the Messages application programming interface and a separate managed-agents layer with stateful sessions, reusable agent definitions, files, skills, and environments. (platform.claude.com, platform.claude.com) The company is pairing those tools with admin controls aimed at information technology teams, not just individual users. Anthropic’s Admin application programming interface can manage organization members, roles, workspaces, and application programming interface keys, which means onboarding and offboarding can be automated instead of handled by hand in a settings page. (platform.claude.com) The permission system is getting more detailed too. Enterprise plans support custom roles at the group level, Team and Enterprise plans support single sign-on, and Enterprise owners can export audit logs covering the past 180 days. (support.claude.com, support.claude.com, support.claude.com) Anthropic is also teaching people how to use Claude as part of a workflow, not as a one-off prompt box. Its free Claude Code 101 course runs 12 lectures over 1.5 hours and teaches an “Explore → Plan → Code → Commit” rhythm, plus project memory files, custom subagents, and external tool connections. (anthropic.skilljar.com) Even the broader training catalog points in the same direction. Anthropic Academy now bundles application programming interface guides, model documentation, prompt engineering, Model Context Protocol lessons, and an AI fluency course built with Joseph Feller of University College Cork and Rick Dakan of Ringling College. (anthropic.com, anthropic.com) Put together, this looks less like a single chat app and more like a work surface with a model underneath it. Claude now has a browser hand, a coding loop, a managed-agent backend, and the security controls companies usually demand before they let software touch real systems. (claude.com, platform.claude.com, platform.claude.com, support.claude.com)