Your apps can now open in Claude, and Claude can work in them

Anthropic launched an expanded system for interactive connectors that allows third-party partner applications to embed and run directly inside Claude. The rollout brings tools alongside conversations or in full-screen workspaces across the web interface and the Claude desktop application for all subscription tiers, as outlined in Claude Blog's report. The release links conversation threads with Claude Cowork, enabling users to direct Claude to perform edits, create records, and update files inside external software without toggling browser tabs or copying snippets between separate windows.

This coverage details the operational mechanics of embedded application tools, the partner software supported at launch, enterprise administrative controls, and the underlying Model Context Protocol open standards. The functionality is designed for knowledge workers, developers, and workspace administrators seeking unified multi-app workflows. Users receive access as individual accounts update to the new interface or when enterprise workspace administrators enable the unified experience.

01Full tool windows and live workspace edits

The updated system replaces small conversational previews with full application interfaces rendered directly beside the dialogue stream. When users open a connected tool like Confluence or Notion, they access the complete native editor authenticated under their own existing credentials. The interface permits hiding the chat window entirely or expanding the external application into a full-screen view whenever complex tasks demand additional visual area.

Claude interacts with active tools by making live changes directly to selected documents, individual spreadsheet rows, or project files. Users approve every discrete action through the standard connector authorization prompt before modifications execute.

All accepted changes populate instantly within the embedded software interface, letting individuals inspect edits in real time alongside automated generation. Connected services can also launch independently without prompting Claude beforehand, allowing users to initiate AI tasks only when manual edits require assistance.

02Directory tool support and creative integrations

Anthropic opened the integration to a core group of directory tools that includes Asana, Canva, Carta, Dropbox, monday.com, Smartsheet, Vercel, and Webflow. Users can pin their most frequent tools beside artifacts, establishing one-click accessibility within the primary Claude sidebar and the universal creation menu. The directory connections allow team members to route work directly into project boards, financial ledger trackers, and cloud repositories from a single unified panel.

Atlassian users can open Confluence pages beside the prompt window to draft prose and review textual adjustments live on the canvas. Box connections support side-by-side file comparisons before committing AI revisions, while HubSpot users can organize sales pipelines and follow-up activities without leaving Claude. Additional integrations from Figma, Harvey, and Shopify are actively under development.

Atlassian users can open Confluence pages beside the prompt window to draft prose and review textual adjustments live on the canvas. Box connections support side-by-side file comparisons before committing AI revisions, while HubSpot users can organize sales pipelines and follow-up activities without leaving Claude. Additional integrations from Figma, Harvey, and Shopify are actively under development.

03Open MCP Apps architecture for developers

Under the hood, interactive connectors operate on MCP Apps, an open extension to the Model Context Protocol that permits MCP servers to deliver user interfaces into host AI environments. Applications built against the specification remain vendor-neutral, ensuring they can execute inside any compatible AI platform that adopts the open standard. The protocol expands capabilities beyond basic structured text responses by embedding complete interactive digital workspaces.

While previous implementations in Claude limited MCP Apps to static or lightweight card views inside the chat stream, the updated system adds broader extension capabilities including app tools and fully embedded web applications. Support for these extended interface capabilities has entered public developer preview, accompanied by developer documentation. Third-party software engineers can immediately begin constructing custom MCP server extensions to package their proprietary web software for distributed AI client runtimes.

04Administrative permissions and directory setup

Accessing tools begins inside the Claude connector directory, where individual users link authorized services before pinning selected items to sidebars or invoking them directly from conversation inputs. The deployment functions on all plans across both desktop and browser environments once the host account receives the updated interface. The workflow eliminates repetitive context copying between fragmented desktop windows.

Enterprise workspace administrators manage security permissions using established connector access rules. If an enterprise workspace currently permits every tool component under a connector, the expanded in-app capabilities deploy automatically as the platform update reaches the tenant. Workspace administrators retain existing controls to limit, restrict, or shut off individual tool permissions across their organizations at any time to enforce enterprise compliance boundaries.

05Unified workspace companion services

The rollout of full embedded tools follows a sequence of workspace expansions within the platform ecosystem. Anthropic introduced companion native tools including Claude Whiteboards, Claude Sheets, and Claude Data to handle brainstorming sessions, spreadsheet generation, and unified data sharing directly within user sessions. The additions coincide with dedicated personal finance management features designed to query linked accounts for structured financial overviews.

These interactive software environments coordinate alongside Anthropic's advanced model tiers, including Claude Fable 5.2 and Claude Mythos 5.2, developed to tackle complex technical calculations and deep analytical workflows. By uniting third-party embedded apps, internal canvas tools, and expanded reasoning models, the environment centralizes administrative operations and software engineering tasks into an integrated operating workspace.