Anthropic introduced a new capability on Monday that allows Claude users to access interactive applications directly within the chatbot. Reflecting the company's emphasis on business clients, the initial selection focuses on workplace utilities such as Slack, Canva, Figma, Box, and Clay, with a Salesforce integration planned for the near future. Once enabled, these apps create a logged-in session that Claude can utilize, letting users perform actions like sending Slack messages, creating visual content, or retrieving files from cloud storage.

"Tasks like data analysis, content creation, and project management benefit significantly from a specialized visual interface," Anthropic explained in its announcement. "When paired with Claude's reasoning, this combination accelerates workflow and iteration beyond what either could achieve independently."

Access to this feature is limited to Pro, Max, Team, and Enterprise subscribers, excluding those on the free tier. Eligible users can enable the tools by visiting claude.ai/directory. This system resembles the Apps platform launched by OpenAI in October, which also supports interactive third-party tools. Both integration frameworks are constructed on the Model Context Protocol (MCP), an open standard Anthropic released in 2024. MCP added support for applications in November, incorporating developments from both organizations.

These new apps are expected to be especially effective when used alongside Claude Cowork, a versatile agent tool Anthropic released last week. Developed from Claude Code, Cowork enables users to delegate complex, multi-step tasks that involve extensive and open-ended datasets—work that traditionally required command-line operations. When integrated with the new apps, Cowork could be authorized to access cloud files or active projects. For instance, it might modify a marketing asset in Figma or process updated information from a company's Box storage.

Although apps are not initially available within Cowork, Anthropic confirmed this functionality is "coming soon." The company acknowledges that agentic systems can behave unpredictably and advises caution in its safety guidelines for Cowork, recommending users supervise the agent actively and limit permissions strictly. "Avoid providing access to sensitive materials such as financial documents, login credentials, or personal records," the guidance states. "It is advisable to establish a designated working directory for Claude instead of allowing extensive system access."