Anthropic integrates third‑party apps into Claude, reshaping enterprise AI workflows

Anthropic has added a new capability inside its generative AI-based chatbot Claude that will allow users to directly access applications inside the Claude interface.

The new capability, termed as interactive apps, is based on a new extension — MCP Apps — of the open source Model Context Protocol (MCP).

MCP Apps, first proposed in November and subsequently developed with the help of OpenAI’s Apps SDK, expands MCP’s capabilities to allow tools to be included as interactive UI components directly inside the chat window as part of a conversation, instead of just text or query results.

“Claude already connects to your tools and takes actions on your behalf. Now those tools show up right in the conversation, so you can see what’s happening and collaborate in real time,” the company wrote in a blog post.

Currently, the list of interactive apps is limited to Amplitude, Asana, Box, Canva, Clay, Figma, Hex, monday.com, and Slack. Agentforce 360 from Salesforce will be added soon, the company said, adding that the list will be expanded to include other applications.

Easier integration into workflows

Claude’s evolution from a chatbot into an integrated execution environment for applications is expected to help enterprises move agentic AI systems toward broader, production-scale deployments, analysts say.

The ability to access applications directly within Claude’s interface lowers integration friction, making it simpler for enterprises to deploy Claude as an agentic system across workflows, said Akshat Tyagi, associate practice leader at HFS Research.

“Most pilots fail in production because agents are unpredictable, hard to govern, and difficult to integrate into real workflows. Claude’s interactive apps change that,” Tyagi noted.

For enterprise developers, the reduction in integration complexity could also translate into faster iteration cycles and higher productivity, according to Forrester Principal Analyst Charlie Dai.

The approach provides a more straightforward path to building multi-step, output-producing workflows without the need for extensive setup or custom plumbing, Dai said.

Targeting more productivity

According to Tyagi, productivity gains will not only be limited to developers, but business teams will stand to benefit as well, and teams don’t need to move between systems, copy outputs, or translate AI responses into actions due to the integration of multiple applications within Claude’s interface.

MCP Apps and Anthropic’s broader approach to productivity underscore a widening architectural split in the AI landscape, according to Avasant research director Chandrika Dutt, even as vendors pursue the same underlying goal of boosting productivity by embedding agents directly into active workflows.

While Anthropic and OpenAI are building models in which applications run inside the AI interface, other big tech vendors, including Microsoft and Google, are focused on embedding AI directly into their productivity suites, such as Microsoft 365 Copilot and Google Gemini Workspace, Dutt said.

“These strategies represent two paths toward a similar end state. As enterprise demand for agent-driven execution grows, in a separate layer, it is likely that these approaches will converge, with Microsoft and Google eventually supporting more interactive, app-level execution models as well,” Dutt added.

Further, the analyst pointed out that Claude’s new interactive apps will also facilitate easier governance and trust — key facets of scaling agentic AI systems in an enterprise.

“Claude operates on the same live screen, data, and configuration the user is viewing, allowing users to see exactly what changes are made, where they are applied, and how they affect tasks, files, or design elements in real time, without having to cross-check across different tools,” Dutt said.

Increased burden of due diligence

However, analysts cautioned that the unified nature of the interface may exponentially increase the risk from a security standpoint, especially as more applications are added.

More so because running UIs of these interactive applications means running code that enterprises didn’t write themselves and that will increase the burden of due diligence before connecting to an interactive application, said Abhishek Sengupta, practice director at Everest Group.

MCP Apps itself, though, offers several security features, in the form of sandboxing UIs, the ability for enterprises to review all templates before rendering, and audit messages between the application server and their Claude client.

The interactive app feature is currently available to all paid subscribers of Claude, including Pro, Max, Team, and Enterprise subscribers.

It is expected to be added to Claude Cowork soon, the company said.