Anthropic just turned Claude into something closer to an operating system. The AI company rolled out MCP Apps today, letting users interact with Slack, Figma, Canva, and Asana directly inside the chatbot without toggling between tabs. It's a sharp departure from the old text-only responses, and it signals how AI assistants are evolving from standalone tools into full-fledged platforms where work actually happens.
Anthropic just redefined what an AI assistant can do. The company announced today that Claude users can now interact with apps like Slack, Figma, Canva, and Asana directly inside the chatbot, eliminating the need to copy-paste text or switch tabs. The new MCP Apps extension transforms Claude from a text-based assistant into an interactive workspace where users can draft Slack messages with full formatting, customize Canva presentations in real time, and manage project timelines in Asana—all without leaving the chat interface.
The shift marks a fundamental change in how enterprise workers might use AI tools. Previously, connecting Claude to apps like Slack or Asana meant getting plain text summaries back. Now, according to Anthropic's announcement, users can "see, explore, and refine results visually, not just read about them." The company says tools will "open as interactive apps right inside of chat," creating what feels like mini-applications embedded within the AI assistant.
The launch comes with immediate availability for several popular enterprise tools. Users can already format and preview Slack messages, build interactive charts using Hex or Amplitude, manage projects with Asana or monday.com, and work with Figma, Clay, and Box integrations. Salesforce tools including Data 360, Agentforce, and Customer 360 apps are "coming soon," Anthropic confirmed, suggesting broader enterprise adoption is on the horizon.
The integration resembles the "mini" apps embedded inside messaging platforms like Telegram and Discord, but it's really about something bigger—AI assistants becoming operating systems. ChatGPT already moved in this direction with its own app ecosystem launch last year, and now Claude is following suit. It's the "everything app" model pioneered by Tencent's WeChat in China, where a single platform becomes the hub for countless activities.
Behind the scenes, Anthropic credits the Model Context Protocol for making this possible. MCP, the open-source protocol that allows AI agents to access tools and data across the internet, now has a new extension called MCP Apps. Crucially, the extension "lets any MCP server deliver an interactive interface within any supporting AI product—not just Claude," meaning this kind of app-within-an-app experience could soon appear in competing AI tools.
Operating standards like MCP matter because they prevent the tech industry from fracturing into incompatible silos. Instead of every company building custom interfaces for every AI platform, MCP creates a common language. The protocol started at Anthropic in 2024 but quickly gained traction with OpenAI, Google, and Microsoft all adopting it. Last year, Anthropic donated MCP to the Linux Foundation and established the Agentic AI Foundation alongside OpenAI, Google, Microsoft, AWS, Block, Bloomberg, and Cloudflare to "advance open-source agentic AI."
The timing reflects broader momentum around AI agents that can actually do things, not just answer questions. Enterprise software companies have been racing to embed AI capabilities, but most integrations still feel clunky—requiring users to manually transfer information between systems. MCP Apps solves that by letting the AI assistant become the interface itself, with third-party tools rendering their own interactive components inside the chat window.
For productivity software companies, this represents both opportunity and threat. On one hand, integration with Claude puts their tools in front of millions of AI users. On the other, it raises questions about who owns the customer relationship when work happens inside an AI assistant rather than the native app. Canva, Figma, and Slack are essentially ceding interface control to Anthropic, betting that convenience outweighs brand presence.
The enterprise angle here is significant. While consumer AI tools grab headlines with flashy demos, the real money is in B2B software subscriptions. If Claude can become the single interface where knowledge workers draft presentations, send messages, analyze data, and manage projects, it fundamentally changes the SaaS landscape. Microsoft clearly sees this coming, which is why it's pushing Copilot across its entire product suite.
What's less clear is how these embedded apps will handle complex workflows that typically require multiple steps across different tools. The demos show relatively simple tasks—drafting a message, customizing a chart, updating a project status. Real work often involves conditional logic, approvals, and integrations between systems. Whether MCP Apps can handle that complexity or if it's limited to surface-level interactions remains to be seen.
Anthropic's MCP Apps launch signals a major shift in how AI assistants will function in enterprise environments. By embedding interactive app interfaces directly into Claude, the company is betting that knowledge workers want a single hub for all their tools rather than juggling multiple browser tabs. With Salesforce integrations on the way and the underlying MCP protocol open to all AI platforms, we're watching the early stages of AI assistants evolving into full-fledged operating systems. The question now isn't whether other platforms will follow—ChatGPT already started down this path—but how quickly enterprise software companies will adapt to a world where their apps live inside someone else's interface.
