Claude users will now be able to call up interactive apps within the chatbot interface, thanks to a new feature announced by Anthropic on Monday.

In keeping with Anthropic’s enterprise focus, the launch apps are mostly workplace tools, including Slack, Canva, Figma, Box, and Clay, with a Salesforce implementation expected soon. In each case, the app will enable a logged-in instance of the service that’s accessible to Claude, enabling users to send Slack messages, generate charts, or access cloud files, depending on which apps have been enabled.

“Analyzing data, designing content, and managing projects all work better with a dedicated visual interface,” Anthropic said in a blog post announcing the feature. “Combined with Claude’s intelligence, you can work and iterate faster than either could offer alone.”

The new feature is available to Pro, Max, Team, and Enterprise subscribers, but not free users. Eligible users can activate the tools at claude.ai/directory.

The system is similar to OpenAI’s Apps system, which launched in October and enables interactive third-party tools. Both systems of app integrations are built on the Model Context Protocol (MCP), an open standard introduced by Anthropic in 2024. MCP launched support for apps in November, drawing on work from both companies.

The new apps will be particularly powerful when integrated with Claude Cowork, an all-purpose agent tool launched by Anthropic last week. Built on top of Claude Code, Cowork lets users assign multistage tasks that draw on large and open-ended datasets — tasks that would have previously required terminal commands. Combined with the new apps feature, Cowork could be granted access to cloud files or ongoing work projects. For example, Cowork could update a marketing graphic in Figma or use new data from the company’s Box instance.

Apps are not available in Cowork at launch, but Anthropic said the integration would be “coming soon.”

Agentic systems can be unpredictable, and Anthropic’s own safety documentation for Cowork encourages users to monitor the agent closely and not grant any unnecessary permissions.

“Be cautious about granting access to sensitive information like financial documents, credentials, or personal records,” the company recommends. “Consider creating a dedicated working folder for Claude rather than granting broad access.”