Isolated Agents SDK

The Isolated Agents SDK is a powerful security framework designed to solve one of the most pressing challenges in the modern AI ecosystem: the safe execution of autonomous agents on local hardware. As AI agents gain the ability to generate and execute code, browse the web, and interact with file systems, the risk of "prompt injection" or malfunctioning logic leading to system compromise has increased significantly. Our SDK provides a "security-first" abstraction layer that automatically wraps any Python-based AI agent—whether built with LangChain, AutoGPT, or custom logic—within a rootless Podman container. This ensures that the agent operates in a completely isolated environment with zero access to the host system unless explicitly granted. Key features include: Automatic Isolation: Move from raw Python execution to a secure container with a single function call (run_agent). Granular Policies: Define strict resource limits (CPU/Memory), network whitelists (e.g., only allow access to specific LLM APIs), and read-only file system access. Seamless Serialization: Uses advanced pickling techniques to transport complex agent states into the container without requiring manual Dockerfile management. Artifact Management: Automatically collect and map generated files, logs, and data back to the host system. By bridging the gap between high-level AI orchestration and low-level system security, the Isolated Agents SDK empowers developers to build and deploy innovative AI tools with total peace of mind.