SAN FRANCISCO, January 26, 2026, 11:33 PST
- Anthropic says Claude can now open interactive versions of workplace tools like Slack, Asana, Canva and Figma inside chat
- The rollout is built on MCP Apps, a new extension to the Model Context Protocol, with broader client support expected
- The shift tightens the link between AI assistants and real workplace systems, raising control and security questions
Anthropic on Monday rolled out interactive versions of workplace tools inside its Claude service, letting users draft and send Slack messages, update Asana timelines and tweak Canva decks without leaving the chat window. The company said the first set includes Slack, Asana, Canva, Figma, Box, Clay, monday.com, Amplitude and Hex, available on web and desktop for paid plans, with Salesforce’s Agentforce 360 coming soon. (Claude)
The move matters because AI firms are trying to turn assistants into places where work gets done, not just discussed. Swapping copy-paste summaries for direct edits is a bigger step, and companies tend to notice once an assistant can touch real messages, tasks and files.
That shift is powered by MCP Apps, billed as the first official extension to the Model Context Protocol (MCP), an open standard meant to connect AI products to external tools and data. The MCP project’s core maintainers said MCP Apps lets tools return interactive user interfaces — dashboards, forms and visualizations — that render in a sandboxed iframe, a locked-down web frame inside the conversation. They said Claude supports it today and ChatGPT is due to add support this week; Anthropic’s David Soria Parra said, “I am excited about the possibilities that MCP Apps opens up,” while OpenAI’s Nick Cooper said, “We’re proud to support this new open standard and look forward to seeing what developers build with it.” (Model Context Protocol Blog)
TechCrunch said the embedded apps tap a logged-in service instance, allowing Claude to post messages or pull cloud files depending on the permissions users grant. (TechCrunch)
The Verge said some connectors already existed, but they mostly returned text that users then had to paste back into the target app. With the new approach, Anthropic said tools will “open as interactive apps right inside of chat,” letting users “see, explore, and refine results visually, not just read about them.” (The Verge)
In practice, this is less about a smarter reply and more about a tighter loop: write, preview, adjust, send — without leaving the thread. For teams already drowning in tabs, that’s the pitch.
It also puts the standards fight in the foreground. If MCP Apps works across clients, developers can build one interactive tool and ship it to multiple assistants, instead of rewriting the same integration for each platform.
But deeper access cuts both ways. A bad draft stays a bad draft until someone hits “post,” and companies will have to decide how much autonomy and access they can live with, especially in tools that hold customer data and internal chats.
Anthropic’s bet is that interactive panels inside chat will feel normal fast, the way file previews and inline docs did before them. The harder part is what happens after the demo: permissions, policies, and whether workplaces let these connectors spread.
