Anthropic PBC recently unveiled a new tool called Claude for Life Sciences aimed at advancing scientific research. In addition, the company expanded the application of its Claude Code agent coding tool to web and mobile platforms. Among these two developments, Claude for Life Sciences appears more promising as it represents a tailored version of Anthropic's most advanced AI model, designed to assist scientists in tasks like drug discovery more efficiently. The company has re-engineered Claude to interface with popular lab tools frequently used by laboratory researchers. This announcement marks the first time Anthropic has entered the field of life sciences. Claude for Life Sciences is intended to support researchers throughout every phase of drug discovery, from reading research to developing theories, data analysis, and even preparing documentation for regulatory submissions. Eric Kauderer-Abrams, head of biology and life sciences at Anthropic, told CNBC that this launch represents a significant milestone in the company's substantial investment in this domain. He added, “We envision a meaningful portion of all life science work globally being conducted on Claude, much like coding is today.” To achieve this, Anthropic has developed several integrations that enable Claude for Life Sciences to work seamlessly with research platforms like Benchling, 10x Genomics, PubMed, and Synapse.org. These integrations allow scientists to directly extract data from these platforms into Claude without exporting files or switching applications. The company has also engaged consulting partners such as Deloitte, KPMG, and Caylent Inc. to help life science organizations make the most of Claude. Kauderer-Abrams, who joined Anthropic a few months ago, stated that the company was inspired by the fact that many researchers were already using the standard Claude model to aid their work. This motivated the creation of a dedicated version of Claude for researchers, along with the necessary infrastructure. Built on Anthropic's most powerful large language model, Claude Sonnet 4.5, Claude for Life Sciences is designed to better perform tasks like reading lab protocols and other science-related activities. During a demonstration, the company showed how a scientist could use Claude for Life Sciences to compare two new drug dosage schedules. She could pull data directly from Benchling, auto-generate tables to compare differences, and then update the information within the original system. Once completed, she used Claude for Life Sciences to generate a regulatory research report. The company explained that such work typically takes days due to the manual effort required to compile results and verify everything, but with Claude, the same can be accomplished in minutes. However, Kauderer-Abrams quickly pointed out that AI cannot accelerate the entire drug discovery process—for example, clinical trials cannot be sped up by AI. Anthropic and Claude aim to expedite the more repetitive aspects of scientific workflows, such as tedious analyses and comparisons, and regulatory documentation. “We are here to ensure this transformation happens—and happens responsibly,” Kauderer-Abrams emphasized.

Claude Code Now Available on Web and Mobile

The second update from Anthropic involves Claude Code, previously limited to developer terminals, now coming to its mobile app and the web application at Claude.ai. The agent will be launched on these platforms as a “research preview,” accessible only to users with Pro and Max accounts, the company noted. This is essentially a new way for developers to interact with Claude Code, which Anthropic believes will be particularly useful for those looking to run multiple clearly defined coding tasks in parallel. Developers can connect the agent to their GitHub repositories, tell it what to do, and then simply monitor the work as new code appears in the web interface. The agent will run in a sidebar while the task list appears on the left. The company emphasized that each Claude Code session runs in its own sandbox environment for security, and all Git interactions go through a secure proxy service, ensuring it only accesses repositories it is authorized to access. Claude Code recently gained a new feature that allows developers to guide it while it works on problems—a feature now available on web and mobile apps, enabling developers to guide the agent without interrupting its work, potentially avoiding the need to restart tasks. “Developers can now manage a group of Claude Code instances rather than handling individual coding tasks one at a time, with confidence that they will complete their work securely and independently,” the company said. “This is less about watching Claude work and more about delegating to an entire team—you assign the work, Claude gets it done, and you review the results as each task finishes.” According to Anthropic, the company has been heavily relying on Claude Code itself, with the agent reportedly writing about 90% of its own codebase. This has resulted in a roughly 67% increase in productivity for its engineering team over the past few months. Additionally, Claude Code has generated significant revenue for the company, with an annualized revenue rate exceeding $500 million.