Claude with Claude Cowork built in: skills, connectors, plugins and what each plan includes

Claude is Anthropic's assistant app on web, macOS, Windows, Linux (beta), iOS and Android. On September 16, 2026 Anthropic merged Claude Cowork, its agent mode for knowledge work, into chat, so a single Claude now decides when a request needs multi-step work. Tasks keep running in the cloud after you close your laptop, you can put a task on a schedule, such as "Every weekday at 8 AM, summarize new messages in my team's Slack channels", and Claude Desktop adds a folder on your computer Claude can read and write, a built-in browser that opens sites and fills in forms, and computer use, in beta on Pro and Max.

#What changed when Claude Cowork merged into chat?

The merge reaches Pro and Max first, on web, desktop and mobile. More plans follow, and Enterprise admins get at least 30 days' notice. Existing Cowork tasks, projects, connectors, skills, artifacts and files carry over and open as before. The help centre lists the rough edges: you cannot branch a conversation from an earlier point, "Add from GitHub" is unsupported, search leaves out older Cowork tasks, Dispatch is closed to new users although existing ones continue, and incognito chats open in the previous experience without file and code capabilities. Claude Docs, Claude Slides and Claude Design are in beta on paid plans. For a comparison with a self-hosted agent, see Claude Cowork vs OpenClaw.

#What does each plan include?

Connectors are on every plan, Free included, though Free allows only one custom connector. Anthropic's pages disagree about skills on Free: the help centre says every plan, the platform docs say Pro and above, so check your own Customize menu. Pro costs $20 a month, or $17 billed annually, and adds Claude Code, Claude Design, Slides and Docs, projects and more models. Max starts from $100 a month for higher limits and priority access at peak times. Plugins are the paid-only piece: the help centre limits them to Pro, Max, Team and Enterprise.

#How do skills, connectors and plugins differ?

A skill is a folder of instructions you zip and upload. A custom connector is the remote MCP server URL of a service, reached from Anthropic's cloud, so a server on a private network needs Anthropic's IP ranges allowlisted in your firewall. A plugin bundles skills, connectors, sub-agents and hooks into one package, and its .claude-plugin/plugin.json manifest is optional: components are auto-discovered from conventional folders. Anthropic's pages conflict on the details: the help centre calls the file skill.md and caps the description at 200 characters, the platform docs call it SKILL.md and allow 1,024. More on each piece in Agent Skills, MCP and Claude Code.