Anthropic lets Claude access private systems

Anthropic's new features allow Claude AI agents to securely access and operate within an enterprise's private systems, addressing major security concerns.

Anthropic has introduced two new enterprise-focused capabilities for its Claude Managed Agents platform: self-hosted sandboxes and MCP tunnels. These features are designed to let autonomous AI agents operate securely within a company's private network. The update addresses a common challenge where organizations want to leverage AI agents for internal tasks but are prevented by security policies that prohibit exposing internal systems or execution environments to the public internet. Self-hosted sandboxes allow companies to run the agent's code execution environment on their own infrastructure, while MCP tunnels create a secure connection for the agent to access internal tools and data.

This development is significant for CTOs, security teams, and developers looking to deploy advanced AI solutions. The primary barrier to using autonomous agents in many enterprises has been the security risk of granting external services access to sensitive internal systems. By providing a way for agents to work without leaving the corporate security perimeter, Anthropic makes its platform more attractive to businesses in regulated industries. This could accelerate the adoption of AI agents for automating complex internal workflows, from managing IT infrastructure to analyzing proprietary business data, making Claude a more competitive option for large-scale enterprise deployments.

It removes a major security barrier for enterprises wanting to use AI agents with their internal systems, potentially accelerating adoption in regulated industries.

Enables businesses to securely automate internal workflows using AI agents without exposing their private infrastructure, potentially increasing operational efficiency and making AI adoption feasible for security-conscious organizations.