In software development the boundary between writing code and shaping the interfaces people actually see has always required switching contexts, tools, and often people.
Over the past year that boundary has continued to blur as language models gain more direct access to codebases and visual editing surfaces. One recent step in that direction arrived as an early preview inside Claude Code.
Developers can now type a command such as /design a few options for {feature} from either the desktop app or the command-line interface.
With the command, Claude can examine the existing repository, notes the current interface patterns, and produces several draft mockups presented as artboards or Artifacts.
A user selects one, makes adjustments with the same editor and prompting style previously available only in the separate Claude Design canvas, and then asks Claude to implement the chosen version.
The designs are meant to carry forward into the build step, though at the moment they still need to be saved manually. Updating Claude Code with the usual command is enough to enable the preview.
It is listed as available on Pro, Max, Team, and Enterprise plans.
Earlier work in June had already linked Claude Design and Claude Code through bidirectional commands that could pull design systems from a local codebase or hand finished designs over for implementation; the new slash command simply brings more of that visual workflow into the coding environment itself.
The same week brought another set of changes aimed at everyday productivity rather than interface design.
Claude can now send messages through Gmail and manage files stored in Google Drive once the corresponding connectors are enabled.
In practice a user can ask Claude to reply to an existing thread; the system drafts the response and, after explicit approval, sends it.
The same approval step applies by default when Claude shares, moves, or deletes files, creates folders, or uploads content to Drive.
Users can also search and read mail with natural-language queries, draft messages, forward threads, or retrieve documents, spreadsheets, presentations, PDFs and other common formats from Drive.
Connection happens through the connectors menu inside Claude or the desktop app.
The features are offered on all paid plans.
On Team and Enterprise accounts an organization owner must first enable the connectors before individual users can authenticate with their Google accounts. Actions that change data remain under user control unless an administrator later adjusts the settings to allow certain steps without repeated confirmation.
Taken together the two updates illustrate a consistent pattern: Claude is being given more direct access to the surfaces where design decisions and routine knowledge work already happen, while still requiring human review for the steps that leave the local environment or the user's account.
As these capabilities expand, Claude is gradually moving beyond being a tool that simply generates code or answers questions.
It is becoming something that can participate directly in the workflows surrounding development, design, communication, and file management.
The remaining friction is largely about how much control users are willing to hand over, and how much of that control Anthropic is willing to automate. For now, the human remains in the loop, but the direction is becoming increasingly clear.
