Today Anthropic shipped Claude Code v2.1.84 and expanded its mobile app with Figma, Canva, and Amplitude integrations. On its own, either would be a routine product update. Taken together with the last seven weeks, something else is happening.
According to Product Compass, which published a shipping calendar analysis on March 24, Anthropic released 74 products and features across four surfaces in the 52 days between February 1 and March 24. Claude Code alone accounted for 28 of those releases. Product Compass is a single source for this count, but the public changelog makes it credible.
The question is whether velocity at this rate produces durable capability or just a growing maintenance burden.
Three releases in four days
v2.1.84 is the third Claude Code version to drop since Monday. The changelog is specific: a PowerShell tool for Windows in opt-in preview, a TaskCreated hook, WorktreeCreate hook for HTTP, improved MCP deduplication, transcript search, idle-return prompts for sessions over 75 minutes, and more than 20 bug fixes.
v2.1.83, March 25, added a managed-settings.d drop-in directory for configuration, CwdChanged and FileChanged hook events, agents with initialPrompt frontmatter, and sandbox failsafe settings.
v2.1.82, March 20, brought a --bare flag for scripted -p calls and a --channels permission relay.
None of these are headline features on their own. The pattern they form is: a team systematically filling gaps that power users have been working around. The hook system is expanding. Configuration management is getting more formal. Windows support is catching up. These are the kinds of changes that tend to precede broader enterprise adoption, not follow it.
The sprint in context
The February-March run started with Claude Code Session Sharing on February 3 and Cowork GSuite connectors the next day. On February 17, Anthropic and Figma launched "Code to Canvas," which converts Claude Code output into editable Figma designs, per CNBC. Claude Code Review, a multi-agent pull request analyzer via GitHub integration, launched in research preview for Teams and Enterprise customers on March 9, per TechCrunch.
March 17 brought Claude Dispatch: a persistent Claude Cowork session running on a user's computer, messageable from mobile. Felix Rieseberg, an Anthropic engineer, described it at launch: "One persistent conversation with Claude that runs on your computer. Message it from your phone. Come back to finished work."
Then this week: Auto Mode in research preview (covered separately), two Claude Code versions, and today's mobile tool integrations.
Auto Mode is worth a brief note here because it fits the same pattern. Anthropic's permission model used to offer two choices: approve every tool call, or bypass permissions entirely with --dangerously-skip-permissions. Auto Mode adds a classifier layer that reviews tool calls before execution, blocking mass deletions, data exfiltration, and malicious code runs while letting safer operations through without interruption. It runs on Sonnet 4.6 and Opus 4.6, requires isolated environments per Anthropic's own guidance, and is live for Team plan users.
