Leaning Technologies unveils in-browser Node.js sandboxes for secure AI code execution

Leaning Technologies has introduced BrowserPod for Node.js, the first component of its broader BrowserPod platform, described as a universal execution layer for compute that runs directly inside the browser. The system offers a sandboxed runtime API that allows AI agents to execute untrusted code locally alongside the user, delivering low latency while maintaining strong data security and locality.

The Node.js runtime is aimed at teams developing agent-driven applications, AI coding environments and web-based IDEs, enabling them to run workloads safely in the browser without relying on costly, session-based cloud sandboxes.

As software shifts towards agent-led workflows, the need to execute untrusted code safely has become central. The challenge now focuses less on whether code runs and more on where execution takes place, what resources are accessible and how organisations can manage both risk and infrastructure costs. Cloud sandboxes have typically filled this role, but they introduce operational overhead and broaden the potential exposure of sensitive data.

BrowserPod moves execution to the client side by combining WebAssembly with the browser’s native security model. This approach supports multi-process workloads and networking capabilities required for production use while keeping execution contained within the browser.

The Node.js engine operates inside an isolated environment featuring a virtual file system and tightly controlled access to system functions. Existing Node tools and applications can run without modification while remaining confined to the browser sandbox.

A key feature, Portals, provides controlled networking that allows services running within a pod to be shared through secure URLs. This enables live previews, collaborative debugging and interactive demonstrations without deploying separate backend infrastructure for each user session.

BrowserPod has been designed for use cases such as agentic development tools, MCP-style integrations, browser-based IDEs and interactive documentation. By shifting compute to the client, it reduces pressure on cloud resources, improves responsiveness and keeps sensitive inputs and outputs closer to the user.

Stefano De Rossi, founder and CEO of Leaning Technologies, said, “Agentic software changes the economics and the security model of code execution. BrowserPod is built as a language-agnostic execution layer for browser-based compute — so teams can run untrusted code where the user already is, with lower latency and less dependence on per-session cloud sandboxes. BrowserPod for Node.js is the first engine, and it sets the foundation for broader runtime support.”

Alessandro Pignotti, founder and CTO of Leaning Technologies, said, “AI products increasingly need a safe place to execute code — not just generate it. BrowserPod provides a sandboxed runtime API with the primitives real products need: process isolation, filesystem, and controlled networking through Portals. Node.js is our first engine, and we’re expanding the platform to support additional runtimes and more complete Linux-class workloads over time.”

Leave a Reply

Your email address will not be published. Required fields are marked *