Anthropic Launches Labs to Incubate Experimental Agentic Tools Led by Mike Krieger

Anthropic announced the launch of Labs on January 13, 2026, a new internal division dedicated to incubating experimental AI products at the frontier of Claude’s capabilities. Mike Krieger, the co-founder of Instagram and former Chief Product Officer, will co-lead this initiative alongside Ben Mann to focus on rapid prototyping and user testing. The move allows the company to explore unpolished versions of tools with early users before scaling them into permanent product offerings.

Ami Vora, who joined the company in late 2025, will now lead the primary Product organization to support the growth of existing Claude experiences. This restructuring separates the high-velocity discovery motion from the stable scaling motion required for enterprise customers. By organizing this way, Anthropic aims to maintain its rapid pace of advancement while ensuring core services remain dependable.

A primary focus of the new Labs ecosystem is the expansion of agentic capabilities, specifically through the Claude Cowork research preview. This desktop application is designed for general productivity and represents a shift toward making AI accessible to non-technical knowledge workers. It features a graphical interface that allows Claude to interact directly with local files and folders for tasks like drafting documents and organizing data.

Claude Cowork is built to function as a digital partner that can perform multi-step tasks autonomously on a user's computer. It differs from developer-centric tools by focusing on everyday office workflows such as processing spreadsheets or synthesizing notes from different directories. This research preview is a key example of the frontier work Labs will prioritize as AI moves closer to full desktop integration.

Alongside Cowork, Labs is advancing tools like Skills and the Claude in Chrome extension to enhance how the model interacts with the web. The Skills system allows organizations to package proprietary workflows into modular folders that Claude can execute on demand. Meanwhile, the browser extension enables the model to see and interact with web pages, supporting a build-test-verify loop for various digital tasks.

These experimental motions follow the significant success of Claude Code, which grew from a research preview to a billion-dollar product in just six months. The company recently acquired the JavaScript runtime Bun to serve as the high-performance infrastructure for these agentic tools. This acquisition underscores the technical depth required to make autonomous AI tools fast and efficient enough for professional use. Anthropic is also balancing this innovation with a strong focus on transparency and public-sector partnerships. The company recently released its compliance framework for California's Transparency in Frontier AI Act and partnered with the US Department of Energy for the Genesis Mission. These efforts ensure that as Labs pushes the boundaries of what AI can do, the company remains aligned with safety and regulatory standards.