Anthropic has released two new security features for its Claude Managed Agents, aimed at increasing enterprise adoption by allowing agents to operate within corporate networks. Self-hosted Sandboxes, now in public beta, enable tool execution to occur on customer infrastructure, with options for self-hosting or using four certified providers like Cloudflare and Vercel. MCP Tunnels, currently in research preview, allow agents to securely connect to internal systems via a single outbound encrypted connection, eliminating the need for inbound firewall rules and keeping sensitive data within the company's network. AI
Summary written by gemini-2.5-flash-lite from 3 sources. How we write summaries →
IMPACT Enables wider enterprise adoption of AI agents by addressing critical security and compliance concerns for sensitive industries.
RANK_REASON Product launch of new security features for an enterprise AI agent product.