Anthropic has updated its Claude Managed Agents ecosystem with a self-hosted sandbox feature, giving organizations full control over their infrastructure when deploying coding agents.
Developments
According to newly released documentation, the self-hosted sandbox feature allows enterprises to run Claude's execution environment on their own infrastructure instead of Anthropic's cloud, meeting strict security compliance standards. Alongside this is the 'claude-api' toolset built into Claude Code, enabling agents to interact with internal APIs more seamlessly. Anthropic also published a detailed blog post guiding users on how to configure and maximize the potential of these managed agents in real-world software development workflows.
Why It Matters
Supporting self-hosted sandboxes provides an answer to data security challenges for enterprises when adopting AI agents into their internal codebases. Instead of sending code to the cloud, agents can now work within the organization's on-premises infrastructure, mitigating the risk of source code leaks. Furthermore, bundling 'claude-api' into Claude Code shows Anthropic's ambition to build a more integrated and powerful ecosystem of AI programming tools compared to its competitors.