Anthropic announced a new feature on Monday allowing Claude users to access interactive apps directly within the chatbot interface. The initial apps target workplace tools such as Slack, Canva, Figma, Box, and Clay, with Salesforce integration planned soon.
These apps provide logged-in instances of each service that Claude can utilize. Users enable Claude to perform service-specific actions, including sending messages in Slack, generating charts in Canva, editing designs in Figma, retrieving files from Box, or managing data in Clay. This setup permits seamless interaction without switching between applications.
Anthropic explained the benefits in its blog post: “Analyzing data, designing content, and managing projects all work better with a dedicated visual interface.” The post continued, “Combined with Claude’s intelligence, you can work and iterate faster than either could offer alone.” Such integration supports quicker workflows by leveraging visual tools alongside Claude’s processing capabilities.
Access to the feature requires a paid subscription. It is available exclusively to Pro, Max, Team, and Enterprise users, excluding free tier subscribers. Eligible users activate the apps through the directory at claude.ai/directory, where they select and authorize specific tools.
The implementation resembles OpenAI’s Apps platform, which launched in October to enable interactive third-party tools within its chatbot. Both platforms rely on the Model Context Protocol (MCP), an open standard that Anthropic introduced in 2024. MCP added support for apps in November, incorporating contributions from Anthropic and OpenAI engineers.
The apps gain enhanced functionality when paired with Claude Cowork, an all-purpose agent tool that Anthropic released the previous week. Cowork builds on Claude Code and allows users to delegate multistage tasks involving large or open-ended datasets. Previously, such tasks demanded manual terminal commands for execution.
Integration with the new apps equips Cowork to handle cloud-based operations and ongoing projects. For instance, Cowork can update a marketing graphic directly in Figma or extract new data from a company’s Box instance. At launch, apps remain unavailable within Cowork, though Anthropic stated the integration would arrive “coming soon.”
Anthropic’s safety documentation for Cowork stresses vigilance in deployment. Users must monitor the agent’s activities closely and limit permissions to essentials. The company advises, “Be cautious about granting access to sensitive information like financial documents, credentials, or personal records.” It further recommends, “Consider creating a dedicated working folder for Claude rather than granting broad access.” These measures mitigate risks from agent unpredictability.
