Claude lets you run Slack and Figma right inside your chat

If your daily workday feels like an endless loop of jumping between Slack, Figma, Asana, and dozens of browser tabs, Anthropic is trying to consolidate your screen.

Anthropic just launched interactive connectors in Claude, powered by an extension to its Model Context Protocol called MCP Apps. Instead of just taking actions in the dark through background APIs, Claude can now render interactive visual interfaces directly inside the chat window.

Users can now open tools and manipulate them on the fly. You can turn a project discussion into an Asana timeline, draft and preview formatted Slack messages before firing them off, or generate visual flowcharts and Gantt charts in FigJam without leaving the conversation.

A mini workspace inside your chat

The initial rollout targets key workplace software. Beyond Slack, Asana, and Figma, the lineup includes Amplitude, Box, Canva, Clay, Hex, and monday.com. A deep integration with Salesforce's Agentforce 360 is also slated to arrive soon.

With these connectors enabled, you can prospect accounts and pull contact details in Clay, preview Box documents inline while extracting takeaways, or explore interactive data dashboards in Hex.

There is, however, some conflicting information about availability. Anthropic's official blog notes that interactive connectors are available across Free, Pro, Max, Team, and Enterprise plans, as well as on Claude Cowork. TechCrunch reported that the tools are limited to paying subscribers and that Cowork support is still coming soon.

The underlying engine is MCP, an open standard Anthropic introduced in 2024 to connect AI models to external tools. OpenAI rolled out its own Apps system on MCP last fall. Because MCP Apps is an open extension, any AI interface that supports the standard can theoretically host these interactive components.

More agent power means tighter controls

The real power of these apps will likely show up inside Claude Cowork, the broad agentic tool Anthropic launched last week on top of Claude Code. Cowork can tackle multi-stage, open-ended tasks that previously required command-line inputs.

Once apps plug into Cowork, an autonomous agent could directly pull reference files from Box or tweak assets in Figma. But autonomous software can still act unpredictably.

Anthropic explicitly warns against granting broad access. Its safety guidelines advise against exposing financial documents, credentials, or personal records to the system, recommending instead that users set up isolated, dedicated folders for Claude to work inside.

Why it matters

We think Anthropic made a sharp product move here by accepting the limits of text-only chat. Plain text simply falls apart when you need to inspect a project timeline, verify data distributions, or fine-tune layout margins. Bringing lightweight, interactive mini-apps directly into the chat stream bridges the gap between conversational AI and traditional software.

For everyday knowledge workers, drafting a Slack announcement with a true-to-life preview removes the anxiety of AI sending something broken or misformatted. It makes interacting with agents much safer and more tactile.

On the industry level, this is a clear bid to turn Claude into the master operating system for enterprise work. By leveraging open MCP standards, Anthropic is nudging SaaS giants to treat their own apps as interchangeable widgets inside Claude. If employees stop opening individual enterprise portals, software vendors risk being reduced to backend utilities.

The real friction will be enterprise security. Giving an AI agent active write permissions and embedded UI across multiple core work tools is an IT nightmare waiting to happen unless granular access controls prove bulletproof.