Claude Cowork News & Updates

Cowork breaks out of the desktop app. Starting July 7 it is reachable from claude.ai on the web and from the sidebar of the Claude app on iOS and Android, rolling out to Max subscribers first. Sessions now run remotely in beta — saved to your Claude account, they continue when your laptop is closed and follow you across devices. Chat and Cowork also get a unified interface: one place for projects and artifacts across both.

Three weeks after the June 12 suspension, the US government lifted export controls on Fable 5 and Mythos 5 on June 30, and Anthropic redeployed both models globally on July 1 — across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. The redeployment ships with a new safety classifier that blocks the flagged jailbreak in over 99% of cases, rerouting those requests to Opus 4.8. Paid plans get Fable 5 included for a limited window before it moves to usage credits.

Mid-June brought a wave of security hardening. Claude Code now blocks destructive git and infrastructure commands in auto mode unless you explicitly ask for them, and auto mode reaches third-party model providers. Separately, Workload Identity Federation replaces static API keys with short-lived, scoped credentials on the Claude Platform.

Claude Code can now publish its work as an artifact — a live web page at a private URL that refreshes in place as the session runs. Think PR walkthroughs, incident investigations, dashboards, and release checklists, built from the full context of your codebase, connectors, and conversation. In beta for Claude Team and Enterprise orgs.

Three updates land for connectors. IT admins can now provision MCP connectors once through their identity provider and push zero-touch access to every employee on first login. Connectors come to mobile in beta, and a new observability dashboard gives connector developers adoption and health metrics across Claude, Claude Code, and Cowork.

Claude Design gets a major update: direct on-canvas editing, design systems imported from GitHub, and two-way sync with Claude Code. Run /design-sync to pull your design system into your repo or push work back into Claude Design — or type /design in the terminal to open a live canvas without leaving Claude Code. In beta on Pro, Max, Team, and Enterprise.

Anthropic had proposed moving programmatic usage — the Agent SDK, claude -p, Claude Code GitHub Actions, and third-party apps — off subscription limits and into a separate monthly credit pool billed at API rates, starting June 15. The change was paused before it took effect: programmatic usage still draws from your plan as before. Interactive use — claude.ai, Claude Code in the terminal, and Claude Cowork — was never affected and is unchanged.

Anthropic released Claude Fable 5 on June 9 — its most capable widely released model, with always-on adaptive thinking, a 1M-token context window, and 128K output tokens. Three days later, on June 12, a US government export-control directive required Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals; to comply, Anthropic disabled both models for all customers. Other Claude models are unaffected.

Anthropic releases Claude Opus 4.8, its most capable model yet — at the same price as 4.7. It posts the strongest computer-use and browser-agent results Anthropic has tested and is noticeably more honest about the limits of its own work, while Claude Code gains Dynamic Workflows for tackling very large-scale problems.

Claude Managed Agents can now run tool execution on infrastructure you control, while Anthropic keeps the agent loop — orchestration, context management, and error recovery. Self-hosted sandboxes enter public beta, and new MCP tunnels let agents reach private network services through an outbound-only, end-to-end encrypted gateway.

Anthropic published three new videos showcasing Claude Cowork — an overview, plus demos of vertical Managed Agents for Legal and Financial Services. The Getting Started clip is now embedded in the Guide hero, and both industry demos live on dedicated landing pages alongside a new Managed Agents overview hub.

Anthropic releases Claude Opus 4.7, its most capable model yet. Significant gains in software engineering, sharper vision (3× higher image resolution), and a new ability to verify its own outputs before reporting back — at the same price as Opus 4.6.

Dispatch creates a single, persistent conversation between your phone and Claude Desktop. Send tasks from your phone, Claude executes them locally on your computer. Come back to finished work.

MCP servers can now request structured input from users mid-workflow. New worktree.sparsePaths setting for large monorepos, /effort slash command for controlling thinking depth, and PostCompact hook for post-compression actions.

Opus 4.6's 1M token context window is now enabled by default for Max, Team, and Enterprise subscribers. Previously opt-in, the expanded context is now the standard experience.

Multiple Windows-specific fixes improve support for international text, clipboard operations, and voice input on Windows and WSL environments.

A series of Claude Code releases bring higher default output tokens, new commands for recurring tasks, faster startup, and significant cost reductions for prompt caching.

Claude's new Code Review feature deploys multiple AI agents to review pull requests in parallel. Agents cross-verify findings and rank issues by severity, achieving a false positive rate below 1%.

Opus 4.6 now defaults to medium effort level for Max and Team subscribers, balancing speed and quality. The 'ultrathink' keyword enables high effort for the next turn when deeper reasoning is needed.

Two major Claude Code features: Remote Control lets you continue CLI sessions from any device via phone, tablet, or browser. Auto-memory allows Claude to save useful context automatically for future conversations.

Anthropic makes Claude Memory available to all users for free, including Free tier. A new import tool allows users to migrate conversation history and preferences from ChatGPT and Gemini.

Following the Pentagon controversy and App Store #1 ranking, Claude sees unprecedented growth — daily signups quadruple and active users grow 60% in a single week.

Anthropic launches Claude Code Security for general availability. The reasoning-based vulnerability scanner has found 500+ issues in production open-source code, offering a fundamentally different approach to static analysis.

Claude reaches #1 on the App Store free apps chart, surpassing ChatGPT for the first time. Consumer backlash against OpenAI's Pentagon partnership drives mass migration to Claude.

Anthropic enforces its Acceptable Use Policy by cutting off Pentagon access to Claude after the Department of Defense fails to respond to a compliance ultimatum issued on February 24.

Anthropic discovers unauthorized use of Claude by the Pentagon for military planning applications. CEO Dario Amodei issues a 72-hour compliance ultimatum demanding the DoD cease all non-approved use cases.

Anthropic launches the largest expansion of Cowork's integration capabilities with 13 new enterprise connectors, 10 department-specific plugins, and a Plugin Create platform for custom integrations.

Cowork now supports recurring scheduled tasks. Configure a task once — daily briefings, weekly reports, automated file processing — and Claude runs it on schedule. Uses the /schedule skill or the Scheduled Tasks sidebar.

Claude is now available as add-ins for Microsoft Excel and PowerPoint. Analyze spreadsheets, generate formulas, create presentations, and automate repetitive Office tasks directly within your workflow.

Anthropic launches self-serve purchasing for the Enterprise plan. Organizations can now sign up, configure SSO, and start using Claude Enterprise without going through a sales process.

Claude Sonnet 4.6 is now the default model for Cowork sessions. Sonnet 4.6 offers a strong balance of speed and capability, making everyday Cowork tasks faster while maintaining high-quality output.

A new unified MCP connector brings Microsoft 365 integration to Claude Cowork. Access Word, Excel, PowerPoint, OneDrive, and SharePoint through a single connector, streamlining enterprise productivity workflows.

Claude Cowork launches on Windows with full feature parity with macOS. All capabilities—including plugins, file access, multi-step task execution, and MCP connectors—are now available on both platforms.

Anthropic releases Claude Opus 4.6 with a 1M token context window (beta), adaptive thinking, and significant improvements in agentic task performance and code review capabilities.

Claude Cowork is now available to Team and Enterprise plans in research preview, bringing autonomous desktop AI capabilities to organizational workflows.

Anthropic releases 11 open-source plugins for Claude Cowork, covering role-specific business workflows. From sales pipeline management to HR onboarding, these plugins bring domain expertise to the desktop agent.

Claude Cowork is now available to Pro subscribers ($20/month), expanding access beyond Max-only. This makes agentic desktop AI accessible to a much wider audience.

Anthropic introduces Cowork, an agentic AI assistant that brings Claude's capabilities to your desktop for knowledge work. Unlike traditional chat, Cowork can take on complex, multi-step tasks and execute them autonomously.

Claude gains the ability to read and analyze health and fitness data on mobile platforms, with enterprise-grade healthcare compliance.

Major enhancements to the Skills system, bringing organization-level management and an open ecosystem for third-party integrations.

Claude can now create and edit office documents directly in the Claude Desktop app.

Team and Enterprise plans gain administrative control over desktop extensions with custom extension support.

Enhanced collaboration features with project permissions and sharing capabilities for enterprise users.