Claude gains Live Artifacts and nine creative tools via Anthropic's MCP

Anthropic is turning Claude into something closer to a working surface than a chatbot. The new piece, announced April 28, is a bundle of nine creative-tool connecto(anthropic.com) MCP Apps in January and Live Artifacts in Cowork this month. Put those together, and Claude stops being just a place where you ask questions. It starts looking like a layer that sits inside the tools and data people already use. (anthropic.com) ### What actually shipped? The April 28 launch is a creative-work push. Anthropic said Claude now connects to Ableton, Adobe, Affinity by Can(claude.com)Up, and Splice. Some of these ground Claude in product docs. Others let it generate or modify work through the app’s own interface or API. The common theme is simple: Claude gets closer to the software where designers, musicians, 3D artists, and live-visual teams already spend their time. (anthropic.com) ### Why does MCP matter here? MCP is Anthropic’s Model Context Protocol — an open standard it introduced in Nove(anthropic.com)2026, Anthropic added MCP Apps, which let those connectors show up with live user interfaces inside Claude itself. So this is not just “Claude can call a tool.” It is “the tool can render inside Claude, and you can work with it there.” That is a much bigger product shift. (anthropic.com) ### What are Live Artifacts? Live Artifacts are the clearest example of where this is going. In Claude Cowork, they are persistent HTML dashboards or tra(anthropic.com)r. When you open one, it can pull fresh data from connected apps and local files. It also keeps version history, so the artifact is not a one-off output — it is a living object Claude can keep updating. (support.claude.com) ### Why is that different from old Artifacts? Old Artifacts were already useful — documents, code, visuals, little apps beside the chat. But they wer(anthropic.com)s in Cowork break that link. They live in a dedicated sidebar, can be refreshed with current data, and are meant to persist across tasks. Basically, Claude is moving from “generate me something” to “maintain this thing for me.” (claude.com) ### What can that look like in practice? Anthropic’s own examples are pretty concrete. A live artifact can be a dashboard of open tasks pulled from Asana(support.claude.com)ull requests. Separately, Cowork already has enterprise connectors for Google Workspace apps including Gmail, plus Slack and other business tools through MCP Apps. That means Claude can assemble a view from inboxes, chats, files, and work apps without the user stitching it together manually. (support.claude.com) ### So is Claude becoming BI software? Not exactl(claude.com)opic already showed Hex inside Claude answering data questions with interactive charts and tables, and Live Artifacts extend that idea into persistent dashboards. The difference from classic BI is that the interface is conversational first. You ask for the dashboard in plain English, then Claude builds and maintains it. That lowers the skill barrier a lot. (claude.com) ### What’s the catch for companies? Access control becomes the whole game. If Claude can read Gmail, Slack, local files, and ta(support.claude.com)lumbing for that: role-based access controls, group spend limits, analytics, and OpenTelemetry events for tool calls, files read or modified, and whether actions were manually or automatically approved. In other words, Anthropic knows this only works in big companies if admins can see and limit what Claude is touching. (claude.com) ### Where is this heading? The creative connectors are the headline, but th(claude.com)end for work that used to live across five tabs and three specialist tools. If Anthropic can keep the permissions clean and the outputs trustworthy, this could make AI feel less like a chat window and more like an operating layer for knowledge work. (anthropic.com)