Anthropic unveils Claude Opus 4.6 with agent commands and PowerPoint integration

Anthropic has rolled out an updated version of its flagship AI model, Claude Opus 4.6, just three months after launching Opus 4.5. The standout feature this time is “agent commands,” which let the model split complex tasks among multiple virtual specialists working in parallel while coordinating their efforts.

Instead of a single agent tackling tasks one after another, the work is divided across several agents-each handling a specific part, explained Scott White, head of product at Anthropic, who compared the system to a well-coordinated human team. For example, when building an app, one agent might design the architecture, another writes the code, a third tests functionality, and a fourth prepares documentation-sharing their progress in real time. This feature is currently in beta for Claude Pro subscribers and also accessible via API.

On the technical side, the context window has been expanded to 1 million tokens-roughly 750,000 words or several thousand pages of text. This allows Claude to analyze massive codebases, lengthy legal contracts, or scientific papers without losing track of context. This capability, first introduced in simpler models in the Sonnet line, now powers Anthropic’s top-tier version.

A strategically important update is the direct integration of Claude into Microsoft PowerPoint. Users can now create and edit presentations within PowerPoint’s sidebar using natural language commands like “generate five slides on AI trends for 2026” or “rephrase this text for executives.” Previously, Claude could only produce a separate file that required manual tweaking afterward.

White noted that Claude has evolved from a niche coding tool (Claude Code) into a versatile assistant for professionals across domains-including financial analysts, product managers, researchers, and marketers. “People aren’t just using Claude Code for programming anymore, but as an effective tool for structured analytical work,” he said.