Anthropic has added interactive tool support to its Claude AI platform, a change powered by the open Model Context Protocol (MCP). The update lets users work directly with external applications inside Claude’s interface rather than relying solely on text interactions with connected services.
Interactive tools arrive in Claude
With the update launched January 26, users can open tools such as project management boards, analytics dashboards, design canvases, and messaging platforms inside Claude’s chat interface. Users can view, edit, and interact with content from services including Asana, Box, Canva, Figma, Hex, monday.com, Slack and others without leaving the Claude window. These in-conversation interactive elements show real content from connected services and let users manipulate that content directly.
The goal is to reduce context switching between tools. For example, a security operations team drafting incident messages in Slack or updating task timelines in Asana can do so while keeping Claude in view. Users can preview message drafts, adjust project task assignments, and generate charts with real data all inside a single workflow.
Model Context Protocol at the core
Anthropic first released MCP to the open community in 2024. The idea behind MCP is to provide an interoperable mechanism to expose functions, data, and controls from external systems into conversational and agentic AI workflows.
Interactive tool support in Claude uses an extension to MCP called MCP Apps. This extension enables any MCP server to supply an interactive user interface that can render and accept user interactions directly inside an AI product. Anthropic says MCP Apps will let developers build interactive experiences that work across products that understand the protocol.
What interactive support changes for workflows
Before this update, many AI users relied on simple integration points that passed text or data back and forth between Claude and external systems. For instance, a user might ask Claude for a summary of recent entries in a document store or have it generate a draft email to send via another platform. Those interactions typically required switching to the target app to finalize or execute outputs.
With interactive support, those outputs become a part of the interface. When a user asks to build a chart from analytic data, the chart itself appears inside Claude’s screen and users can adjust parameters in place. Project timelines and boards from collaboration platforms are visible within the conversation and edit controls appear alongside the content.
Anthropic says this integration will be available for web and desktop versions of Claude and in several subscription tiers including Pro, Team and Enterprise. The company also notes that further integrations and tool support are in development.
Developer and ecosystem implications
MCP’s open nature has attracted other AI platform partners and tool providers who see value in a shared specification for exposing data and controls to large models. Industry observers have compared MCP to earlier interface standards that simplified how software components communicate with each other in complex environments.
MCP provides a formal way to build connectors that let AI clients interact with enterprise systems and internal data stores. MCP servers can be custom built to pull from proprietary databases, CRM systems, internal documentation, or operational tools, giving organizations a path to bring internal context to AI-assisted workflows.
Security considerations are part of the conversation around MCP adoption. Security researchers and integrators have noted attack vectors such as prompt injection and permission escalation can affect connectors that expose sensitive systems. Effective use of MCP servers typically involves managing access controls and validating the behavior of connected services.