Claude Agent Threads Live March 17 – Which Model Wins?

Claude's persistent agent thread launched March 17, enabling users to manage Cowork tasks from mobile and desktop for the first time. But as Anthropic adds agentic features weekly, which Claude AI version should you actually adopt right now?

Anthropic's Claude AI Gets Mobile Agent Control

On March 17, 2026, Anthropic announced a significant upgrade to Claude for Pro and Max subscribers: persistent agent threads that let users control Cowork tasks from mobile and desktop without losing context. This is the latest in a series of March releases that reshape how teams use Claude AI for knowledge work and automation.

The persistent thread feature is rolling out first to Max plan users, then Pro users over the following two days. Users can now access a dedicated agent thread via Claude Desktop or Claude for iOS and Android, managing recurring tasks and one-off assignments directly from their phones—a capability that previously required desktop access.

Claude AI's Visual Intelligence Arrives in Chat

Three days earlier, on March 12, Anthropic announced that Claude AI can now create custom charts, diagrams, and inline visualizations directly in responses. This isn't just text-to-code; the visualizations render inside the chat interface itself, meaning users see interactive graphs and design mockups without copying code to external tools.

This addition addresses a core friction point: professionals using Claude for analysis, reporting, or design work previously had to export data and build visuals separately. Now, Claude can generate candlestick charts for financial data, flowcharts for workflows, or UML diagrams for software architecture—all displayed inline.

Enterprise Features Accelerate: Plugins, Integrations, and Controls

Throughout February and into March, Anthropic's Claude AI has gained enterprise-grade capabilities:

• March 11: Claude for Excel and PowerPoint add-ins now share full conversation context across applications, support skills, and connect to Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry via LLM gateway—allowing large organizations to route Claude requests through their existing cloud infrastructure.
• February 24: Anthropic launched a plugin marketplace for Cowork with new admin controls for Team and Enterprise plans.
• February 25: Users can now schedule recurring and on-demand tasks in Cowork, with a new Customize section grouping skills, plugins, and connectors.

Together, these releases position Claude AI as a platform for agentic workflows, not just chat-based assistance.

The Model Upgrades That Matter: Sonnet 4.6 and Opus 4.6

Beneath the feature updates, Anthropic made two foundational model upgrades in February:

Claude Sonnet 4.6 (launched February 17) brings frontier-class performance in coding, computer use, long-context reasoning, agent planning, and design work. The model includes a 1M token context window in beta—equivalent to roughly 700,000 words—enabling teams to work with entire codebases, research libraries, or document sets in a single request. For most organizations, Sonnet 4.6 represents the optimal balance of capability and cost.

Claude Opus 4.6 (launched February 5) is Anthropic's smartest model. It improves coding skills and excels at agentic reasoning, computer use, tool use, search, and financial analysis. Industry benchmarks show it leads by