Beyond AI Sandboxing: Decoupling AI from Execution

AI coding assistants are transforming software development, but they also introduce a new security challenge. Today’s AI agents typically execute directly on developer workstations or require developers to expose secrets, cloud credentials, and production systems to hosted platforms. This creates unnecessary trust assumptions and significantly expands the attack surface, especially when AI-generated applications depend on hundreds of open-source packages that may themselves be compromised. The question organizations should be asking is no longer “Which AI model should we use?” but rather “Where should AI agents execute, and what should they be allowed to access?”

This session introduces a different architectural approach using the open-source Agentry platform. Rather than sandboxing the AI agent itself, Agentry allows any MCP-compatible AI coding assistant—such as Claude Code, Cursor, Cline, or OpenCode—to remotely operate a secure development environment running on infrastructure you already control. Developers continue using their preferred AI tools and can seamlessly switch between different models and coding assistants throughout the development lifecycle without changing the execution environment. Through architecture deep-dives and live demonstrations, attendees will learn how remote sandboxing, zero-trust connectivity, isolated execution, and containerized deployments enable secure AI-assisted software development without disrupting existing developer workflows.

Takeaways

• Understand a new security architecture for AI-assisted development where AI agents interact with remote sandboxed environments over MCP instead of executing directly on developer machines or hosted cloud platforms.

• Learn how remote sandboxing preserves developer productivity by enabling teams to continue using their preferred AI coding assistants while securely building, testing, and deploying applications on infrastructure they already own.

Who will benefit from this session?

This session is ideal for:

• Software Engineers and Full-Stack Developers using AI coding assistants
• Platform Engineers and DevOps Engineers
• Security Engineers and Application Security (AppSec) professionals
• Engineering Managers and Technical Architects evaluating enterprise AI adoption
• Open Source contributors and AI infrastructure enthusiasts

About me

Gowtham Sadasivam is a Senior Staff Engineer at Acceldata with 13+ years of experience designing and operating Linux, Kubernetes, and cloud-native application development. His current work focuses on AI infrastructure, Agentic AI, self-hosted LLMs, and secure software engineering.