AI labs are increasingly looking to provide more AI tools for desktop environments.
About a week ago, ChatGPT introduced Computer History to enable ChatGPT to recall desktop activity. Similarly, Nvidia released an open model (Nemotron 3.5 Lightning) in recent weeks targeting local hardware. The trend toward deploying more AI tools in local systems isn’t stopping.
Anthropic has just announced that the Computer use tool, Skills API, and File API are now publicly available in their developer and enterprise environment, the Claude platform. These features will enable Claude to access a user’s computer, use domain-specific instructions, and to access hosted files in the agentic workflow.
How the new capabilities work
This public release comes a few months after the company announced that Computer use was available in research preview for Cowork and Claude Code users on its paid plans. Now, Computer use is production-ready and can read and write files, take screenshots, and control the mouse or keyboard.
The key addition to Computer use is the new Browser use tool, which now enables it to access the internet for tasks that require the web.
Anthropic also says that it has simplified the Skills API “for uploading and versioning your own skills.” These agent skills enable developers to specify reusable instructions, context, and workflows for well-known tasks.
Furthermore, the File API now features a rate limit that is five times higher and 1TB of storage. This means that developers can access storage via upload or the code execution tool without worrying about hitting the platform limits early.
Safety and security
The new local capabilities Claude has introduced offer more powerful tools for local AI use. But it also raises questions about data and tool safety.
Anthropic has provided guidelines for developers and platform safeguards to follow as they navigate this new landscape. For example, for the Skills API, the company has zero data retention (ZDR), meaning no data is retained on the servers when this tool is called.
For Computer use, the company also shared that the tool is now eligible for Health Insurance Portability and Accountability Act (HIPAA) compliance for enterprises. Despite the safeguards the team has put in place, developers still need to heed the guidelines Claude has shared and do their due diligence.
From the Computer use developer documentation, the company warns that: “In some circumstances, Claude will follow commands found in content even when they conflict with your instructions. For example, instructions on webpages or contained in images might override your instructions or cause Claude to make mistakes.”
What Claude’s new tools mean for the channel
Anthropic’s expanding agent capabilities could create new opportunities for MSPs, systems integrators, and other solution providers as businesses look for ways to put AI agents to work across everyday operations.
Tools that allow Claude to interact with software, browse the web, access files, and follow specialized workflows could increase demand for partners who can help customers design, integrate, and manage agentic systems tailored to specific business needs. That work could extend beyond deployment to include access controls, security, and ongoing governance.
The security considerations may be just as important as the capabilities themselves. As AI agents gain permission to take actions across business systems, channel partners could play a larger role in helping customers determine what those agents should be allowed to access, where human approval is still necessary, and how agent activity should be monitored.
Meanwhile, Anthropic is also scaling the infrastructure behind Claude, reportedly signing a $9.1 billion, 20-year compute deal with Riot Platforms for AI data center capacity in Texas.