Anthropic's latest announcement introduces two significant privacy and security enhancements to Claude Managed Agents, offering users greater control and customization. These updates, MCP tunnels and self-hosted sandboxes, are designed to fortify the security of AI agents while providing a more tailored and flexible deployment experience.
MCP Tunnels: Securing the Agent's Reach
One of the key features, MCP tunnels, enables Claude Managed Agents to securely access internal resources without exposing them to the public internet. This is particularly useful for organizations that want to keep their data and services within a private network. By utilizing a lightweight gateway, MCP tunnels ensure that agents can interact with internal databases, APIs, knowledge bases, and ticketing systems, all while maintaining end-to-end encryption and simplifying firewall management.
The limited research preview status of MCP tunnels means that users need to request access to utilize this feature. This controlled release allows Anthropic to gather feedback and refine the technology before a wider release.
Self-Hosted Sandboxes: Customization and Control
The other new feature, self-hosted sandboxes, empowers users to define their own boundaries for Claude Managed Agents. This means that sensitive files, packages, and services can be managed within the user's infrastructure or with trusted partners like Cloudflare, Daytona, Modal, and Vercel. By keeping the agent loop, which handles orchestration, context management, and error recovery, on Anthropic's infrastructure, while moving tool execution to the user's environment, this feature offers a high level of customization and control.
The self-hosted sandbox feature is available as a public beta, providing users with an opportunity to experiment and provide feedback. This iterative approach ensures that the feature evolves based on real-world usage and user preferences.
Broader Impact and Future Developments
These updates to Claude Managed Agents demonstrate Anthropic's commitment to providing enterprise-grade solutions that are both secure and flexible. By offering MCP tunnels and self-hosted sandboxes, the company is addressing the need for robust security measures and the ability to customize AI agent deployments. This is particularly relevant in industries where data privacy and security are paramount.
As AI continues to integrate into various business processes, the ability to control and secure these agents becomes increasingly important. Anthropic's enhancements in this area not only benefit enterprise users but also contribute to the overall maturity and reliability of AI technologies. The company's focus on security and customization positions it as a leader in the development of AI solutions that meet the evolving needs of businesses and organizations.
